{ "issues": [ { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/314", "id": 3795106690, "node_id": "I_kwDOPT-eoM7iNLuC", "number": 314, "title": "The LTX2 open-source version supports voice-to-video functionality and also supports Chinese. How do you feel about this? Haha", "user": { "login": "2063657", "id": 74194629, "node_id": "MDQ6VXNlcjc0MTk0NjI5", "avatar_url": "https://avatars.githubusercontent.com/u/74194629?v=4", "gravatar_id": "", "url": "https://api.github.com/users/2063657", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2026-01-09T01:25:38Z", "updated_at": "2026-01-09T16:09:38Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/314/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/313", "id": 3793165311, "node_id": "I_kwDOPT-eoM7iFxv_", "number": 313, "title": "目前的视频生成模型都是首尾帧+prompt，是否拥有多轮对话方式(openai chat 的方式)生成视频的，可以参考多张图/视频", "user": { "login": "zx4321", "id": 28834820, "node_id": "MDQ6VXNlcjI4ODM0ODIw", "avatar_url": "https://avatars.githubusercontent.com/u/28834820?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zx4321", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2026-01-08T14:40:04Z", "updated_at": "2026-01-08T14:40:04Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "目前的视频生成模型都是首尾帧+prompt，是否拥有多轮对话方式(openai chat 的方式)生成视频的，可以参考多张图/视频", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/313/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/311", "id": 3787479757, "node_id": "I_kwDOPT-eoM7hwFrN", "number": 311, "title": "https://tongyi.aliyun.com/wan/ 这个体验网站不可用", "user": { "login": "wpr7280", "id": 8286717, "node_id": "MDQ6VXNlcjgyODY3MTc=", "avatar_url": "https://avatars.githubusercontent.com/u/8286717?v=4", "gravatar_id": "", "url": "https://api.github.com/users/wpr7280", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2026-01-07T06:06:28Z", "updated_at": "2026-01-07T06:06:28Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/311/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/309", "id": 3781429716, "node_id": "I_kwDOPT-eoM7hZAnU", "number": 309, "title": "Wan2.2-TI2V中的ref_conv缺少权重和偏置", "user": { "login": "WeorXim", "id": 163127298, "node_id": "U_kgDOCbkgAg", "avatar_url": "https://avatars.githubusercontent.com/u/163127298?v=4", "gravatar_id": "", "url": "https://api.github.com/users/WeorXim", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2026-01-05T13:06:06Z", "updated_at": "2026-01-05T13:06:06Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "设置是这样的，想让传入的图片当作生成**视频的背景**\n\nvideo = pipe(\n prompt=\"A person wearing a helmet and red shorts is riding a bicycle on a paved road next to a grassy area. Realistic video, natural lighting, stable camera, medium shot, eye-level.\",\n negative_prompt=\"色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走\",\n seed=0, tiled=True,\n height=384, width=640,\n reference_image=image, #这里\n num_frames=81,\n)\n\n但是运行报错，显示\nTraceback (most recent call last):\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/examples/wanvideo/model_inference/Wan2.2-TI2V-5B.py\", line 10, in \n pipe = WanVideoPipeline.from_pretrained(\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/diffsynth/pipelines/wan_video.py\", line 129, in from_pretrained\n model_pool = pipe.download_and_load_models(model_configs, vram_limit) #多模型池 支持不同参数模型、不同版本VAE的切换\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/diffsynth/diffusion/base_pipeline.py\", line 289, in download_and_load_models\n model_pool.auto_load_model(\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/diffsynth/models/model_loader.py\", line 70, in auto_load_model\n model = self.load_model_file(config, path, vram_config, vram_limit=vram_limit)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/diffsynth/models/model_loader.py\", line 40, in load_model_file\n model = load_model(\n ^^^^^^^^^^^\n File \"/root/shared-nvme/hf-cache/DiffSynth-Studio/diffsynth/core/loader/model.py\", line 49, in load_model\n model.load_state_dict(state_dict, assign=True)\n File \"/root/shared-nvme/miniconda3/envs/wan/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 2635, in load_state_dict\n raise RuntimeError(\nRuntimeError: Error(s) in loading state_dict for WanModel:\n Missing key(s) in state_dict: \"ref_conv.weight\", \"ref_conv.bias\". \n\n缺少ref_conv的权重和偏置，这是哪里出了问题？？", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/309/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/308", "id": 3779189666, "node_id": "I_kwDOPT-eoM7hQdui", "number": 308, "title": "Both having sex", "user": { "login": "shruti795949-gif", "id": 252774718, "node_id": "U_kgDODxEJPg", "avatar_url": "https://avatars.githubusercontent.com/u/252774718?v=4", "gravatar_id": "", "url": "https://api.github.com/users/shruti795949-gif", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2026-01-04T12:20:53Z", "updated_at": "2026-01-05T08:03:09Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Both having sex", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/308/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/307", "id": 3779189256, "node_id": "I_kwDOPT-eoM7hQdoI", "number": 307, "title": "Both having sex", "user": { "login": "shruti795949-gif", "id": 252774718, "node_id": "U_kgDODxEJPg", "avatar_url": "https://avatars.githubusercontent.com/u/252774718?v=4", "gravatar_id": "", "url": "https://api.github.com/users/shruti795949-gif", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2026-01-04T12:20:21Z", "updated_at": "2026-01-04T12:20:21Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/307/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/306", "id": 3778702370, "node_id": "I_kwDOPT-eoM7hOmwi", "number": 306, "title": "Animate-14B 在 A800 上的计算效率基准", "user": { "login": "Lehyu", "id": 14256558, "node_id": "MDQ6VXNlcjE0MjU2NTU4", "avatar_url": "https://avatars.githubusercontent.com/u/14256558?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Lehyu", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2026-01-04T02:40:36Z", "updated_at": "2026-01-04T02:40:36Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "我在A800上测试了Animate-14B模型替换模式下的视频生成速率，在480P、30fps、20次抽样，其他默认参数的情况下，单次抽样的速率在 10.76s/it 波动，15秒的视频总耗时大概在 10.76s/it * 20 * ceil(15*30/77) = 21.5分钟。\n\n是否有官方基准指标参考下？", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/306/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/304", "id": 3769812773, "node_id": "I_kwDOPT-eoM7gsscl", "number": 304, "title": "Image Address Not Affecting Video Creation in Wan-AI/Wan2.2-I2V-A14B Model", "user": { "login": "majidit", "id": 29058805, "node_id": "MDQ6VXNlcjI5MDU4ODA1", "avatar_url": "https://avatars.githubusercontent.com/u/29058805?v=4", "gravatar_id": "", "url": "https://api.github.com/users/majidit", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2025-12-30T10:56:59Z", "updated_at": "2025-12-31T15:12:19Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": " It appears that the model is not utilizing the provided image addresses when generating videos. Instead, it solely relies on the prompt for video creation.\n\nThe image address provided is direct and does not trigger any CORS (Cross-Origin Resource Sharing) errors.\n\n\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/304/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/303", "id": 3765382181, "node_id": "I_kwDOPT-eoM7gbywl", "number": 303, "title": "ISSUE DECODING IMAGES.", "user": { "login": "parthwagh9999", "id": 59022259, "node_id": "MDQ6VXNlcjU5MDIyMjU5", "avatar_url": "https://avatars.githubusercontent.com/u/59022259?v=4", "gravatar_id": "", "url": "https://api.github.com/users/parthwagh9999", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-28T06:16:31Z", "updated_at": "2025-12-28T06:16:31Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I am facing problem generating images.\n\nI was able to generate images through wan 2.2 when it was launch but haven't used it long time.\n\nTo check above workflow, I ran it again. It was generated 1 image with upscaling as well but, for next, it was stopping at the decode phase. \n\nthis is the workflow: https://www.reddit.com/r/comfyui/comments/1mlvwh1/wan_22_text2image_custom_workflow_v2/?sort=new\n\nTried old workflow as well, same issue. it is reconnecting every time when goes to decode node.\n\nclearly it has generated images in ksampler, but while decoding it is still not creating final image.\n\nI am using Q6, RTX 4090 24GB, wan 2.2 / wan 2.1 also. but same issue.\n\nIs there any issue with coding or cache?\n\nTo resolve this issue, I updated entire comfyui and python dependencies. But still not working.\n\n\"Image\"\n\n\"Image\"\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/303/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/301", "id": 3751710147, "node_id": "I_kwDOPT-eoM7fno3D", "number": 301, "title": "Long waited Wan 2.2 perfect training tutorial has arrived - as low as 6 GB GPUs - 1-click to install, setup, start training, both Windows and Cloud", "user": { "login": "FurkanGozukara", "id": 19240467, "node_id": "MDQ6VXNlcjE5MjQwNDY3", "avatar_url": "https://avatars.githubusercontent.com/u/19240467?v=4", "gravatar_id": "", "url": "https://api.github.com/users/FurkanGozukara", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-21T22:47:53Z", "updated_at": "2025-12-21T22:47:53Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "# Thank you so much Alibaba team for this great open source model. I did massive R&D to prepare this tutorial and presets for end users to directly train locally.\n\n# Wan 2.2 Complete Training Tutorial - Text to Image, Text to Video, Image to Video, Windows & Cloud\n\n## Full tutorial link > https://www.youtube.com/watch?v=ocEkhAsPOs4\n\n[![Wan 2.2 Complete Training Tutorial - Text to Image, Text to Video, Image to Video, Windows & Cloud](https://img.youtube.com/vi/ocEkhAsPOs4/sddefault.jpg)](https://www.youtube.com/watch?v=ocEkhAsPOs4 \"Wan 2.2 Complete Training Tutorial - Text to Image, Text to Video, Image to Video, Windows & Cloud\")\n\nWan 2.2 training is now so easy. I have done over 64 different unique Wan 2.2 trainings to prepare the very best working training configurations for you. The configurations are fully working locally with as low as 6 GB GPUs. So you will be able to train your awesome Wan 2.2 image or video generation LoRAs on your Windows computer with easiness. Moreover, I have shown how to train on cloud platforms RunPod and Massed Compute so even if you have no GPU or you want faster training, you can train on cloud for very cheap prices fully privately.\n\n📂 Resources & Links:\n\nDownload the One-Click Installer & Configs: [ [https://www.patreon.com/posts/Musubi-Tuner-Trainer-App-Configs-137551634](https://www.patreon.com/posts/Musubi-Tuner-Trainer-App-Configs-137551634) ]\n\nQwen Image Model Training Tutorial (Prerequisite): [ [https://youtu.be/DPX3eBTuO_Y](https://youtu.be/DPX3eBTuO_Y) ]\n\nSwarmUI & ComfyUI Setup Guide for Windows: [ [https://youtu.be/c3gEoAyL2IE](https://youtu.be/c3gEoAyL2IE) ]\n\nSwarmUI Installer and Model Downloader : [ [https://www.patreon.com/posts/SwarmUI-Install-Download-Models-114517862](https://www.patreon.com/posts/SwarmUI-Install-Download-Models-114517862) ]\n\nComfyUI Installer : [ [https://www.patreon.com/posts/ComfyUI-Installers-105023709](https://www.patreon.com/posts/ComfyUI-Installers-105023709) ]\n\nSwarmUI & ComfyUI Setup Guide for RunPod & Massed Compute: [ [https://youtu.be/bBxgtVD3ek4](https://youtu.be/bBxgtVD3ek4) ]\n\nUpload / Download Big Files Guide for RunPod & Massed Compute: [ [https://youtu.be/X5WVZ0NMaTg](https://youtu.be/X5WVZ0NMaTg) ]\n\n⏱️ Video Chapters:\n\n[00:00:00](https://youtu.be/ocEkhAsPOs4?t=0) Introduction to Wan 2.2 Training & Capabilities\n- [00:00:56](https://youtu.be/ocEkhAsPOs4?t=56) Installing & Updating Musubi Tuner Locally\n- [00:02:20](https://youtu.be/ocEkhAsPOs4?t=140) Explanation of Optimized Presets & Research Logic\n- [00:04:00](https://youtu.be/ocEkhAsPOs4?t=240) Differences Between T2I, T2V, and I2V Configs\n- [00:05:36](https://youtu.be/ocEkhAsPOs4?t=336) Extracting Files & Running Update Batch File\n- [00:06:14](https://youtu.be/ocEkhAsPOs4?t=374) Downloading Wan 2.2 Training Models via Script\n- [00:07:30](https://youtu.be/ocEkhAsPOs4?t=450) Loading Configs: Selecting GPU & VRAM Options\n- [00:09:33](https://youtu.be/ocEkhAsPOs4?t=573) Using nvitop to Monitor RAM & VRAM Usage\n- [00:10:28](https://youtu.be/ocEkhAsPOs4?t=628) Preparing Image Dataset & Trigger Words\n- [00:11:17](https://youtu.be/ocEkhAsPOs4?t=677) Generating Dataset Config & Resolution Logic\n- [00:12:55](https://youtu.be/ocEkhAsPOs4?t=775) Calculating Epochs & Checkpoint Save Frequency\n- [00:13:40](https://youtu.be/ocEkhAsPOs4?t=820) Troubleshooting: Fixing Missing VAE Path Error\n- [00:15:12](https://youtu.be/ocEkhAsPOs4?t=912) VRAM Cache Behavior & Training Speed Analysis\n- [00:15:51](https://youtu.be/ocEkhAsPOs4?t=951) Trade-offs: Learning Rate vs Resolution vs Epochs\n- [00:16:29](https://youtu.be/ocEkhAsPOs4?t=989) Installing SwarmUI & Updating ComfyUI Backend\n- [00:18:13](https://youtu.be/ocEkhAsPOs4?t=1093) Importing Latest Presets into SwarmUI\n- [00:19:25](https://youtu.be/ocEkhAsPOs4?t=1165) Downloading Inference Models via Script\n- [00:20:33](https://youtu.be/ocEkhAsPOs4?t=1233) Generating Images with Trained Low Noise LoRA\n- [00:22:22](https://youtu.be/ocEkhAsPOs4?t=1342) Upscaling Workflow for High-Fidelity Results\n- [00:24:15](https://youtu.be/ocEkhAsPOs4?t=1455) Increasing Base Resolution to 1280x1280\n- [00:27:26](https://youtu.be/ocEkhAsPOs4?t=1646) Text-to-Video Generation with Lightning LoRA\n- [00:30:12](https://youtu.be/ocEkhAsPOs4?t=1812) Image-to-Video Generation Workflow & Settings\n- [00:31:35](https://youtu.be/ocEkhAsPOs4?t=1895) Restarting Backend to Clear VRAM for Model Switching\n- [00:33:45](https://youtu.be/ocEkhAsPOs4?t=2025) Fixing RAM Crashes with Cache-None Argument\n- [00:35:13](https://youtu.be/ocEkhAsPOs4?t=2113) Dual Model (High & Low Noise) Training Setup\n- [00:36:54](https://youtu.be/ocEkhAsPOs4?t=2214) Preparing Hybrid Datasets (Images + Videos)\n- [00:37:40](https://youtu.be/ocEkhAsPOs4?t=2260) Manually Editing Dataset TOML for Resolution Control\n- [00:39:53](https://youtu.be/ocEkhAsPOs4?t=2393) Setting High Noise Model Paths for Dual Training\n- [00:41:50](https://youtu.be/ocEkhAsPOs4?t=2510) Optimization: Block Swap vs CPU Offload\n- [00:43:10](https://youtu.be/ocEkhAsPOs4?t=2590) Generating Video with Dual-Model Trained LoRA\n- [00:45:35](https://youtu.be/ocEkhAsPOs4?t=2735) Massed Compute: Server Setup & Coupon Code\n- [00:47:00](https://youtu.be/ocEkhAsPOs4?t=2820) Connecting via ThinLinc & File Transfer Methods\n- [00:49:12](https://youtu.be/ocEkhAsPOs4?t=2952) Massed Compute: Fast UV Installation & Downloads\n- [00:50:27](https://youtu.be/ocEkhAsPOs4?t=3027) Loading Configurations on Massed Compute\n- [00:52:18](https://youtu.be/ocEkhAsPOs4?t=3138) Troubleshooting: Fixing Config Version Error\n- [00:53:20](https://youtu.be/ocEkhAsPOs4?t=3200) Dual Model Training Speed Analysis on Cloud\n- [00:55:40](https://youtu.be/ocEkhAsPOs4?t=3340) RunPod: Selecting the Correct Template & GPU\n- [00:57:45](https://youtu.be/ocEkhAsPOs4?t=3465) RunPod: Uploading Files & Extracting Archive\n- [00:58:38](https://youtu.be/ocEkhAsPOs4?t=3518) RunPod: Terminal Installation & Model Downloads\n- [01:00:26](https://youtu.be/ocEkhAsPOs4?t=3626) RunPod: Correct Pathing Syntax & Backslash Fix\n- [01:01:28](https://youtu.be/ocEkhAsPOs4?t=3688) Setting Dataset Paths on RunPod\n- [01:03:34](https://youtu.be/ocEkhAsPOs4?t=3814) Installing nvitop on RunPod Terminal\n- [01:03:54](https://youtu.be/ocEkhAsPOs4?t=3834) Speed Hack: Disabling Numpy Memory Mapping\n- [01:06:00](https://youtu.be/ocEkhAsPOs4?t=3960) Terminating Instances & Final Remarks\n\n\n![Image](https://github.com/user-attachments/assets/1f305387-52f3-4679-8f6d-7f5eb58f76e2)\n![Image](https://github.com/user-attachments/assets/2e342898-e308-4bea-9cae-7daa6ab3c243)\n![Image](https://github.com/user-attachments/assets/5534e0f1-b59d-4d32-951c-73fb70e1c18f)\n\"Image\"\n![Image](https://github.com/user-attachments/assets/f786f0f1-d714-4b7a-b01d-f1c8221fc4a9)\n\"Image\"\n![Image](https://github.com/user-attachments/assets/1341077c-2f40-4d17-935d-5996a2b31d7c)\n![Image](https://github.com/user-attachments/assets/07477334-4400-4656-adc0-dbc36077c1a4)\n![Image](https://github.com/user-attachments/assets/2a798ef1-250d-4239-989a-defd671646b0)\n![Image](https://github.com/user-attachments/assets/8272aedb-876c-4bce-9208-c8ccb2528c69)\n![Image](https://github.com/user-attachments/assets/867d51a3-02c1-476d-b987-b179831a19dd)\n![Image](https://github.com/user-attachments/assets/6eca296a-36e8-4926-acf5-779230acce3e)\n![Image](https://github.com/user-attachments/assets/3dc8b3a9-9144-4396-af16-8a49b06d61d7)\n![Image](https://github.com/user-attachments/assets/8e833d42-af09-425f-9806-b2cf54bd16f3)\n![Image](https://github.com/user-attachments/assets/bfa19197-b69b-415d-a444-796c0db415e7)\n![Image](https://github.com/user-attachments/assets/26aba28f-bdb4-4197-b71c-58b437dee3ef)\n![Image](https://github.com/user-attachments/assets/3020fe7d-df1f-4c62-8307-a278ace57a1a)\n![Image](https://github.com/user-attachments/assets/adedc80d-a48d-4cee-b3b8-6c8c5f80a421)\n![Image](https://github.com/user-attachments/assets/038ed2e5-6c7c-4c88-af27-1790265ff3d4)\n![Image](https://github.com/user-attachments/assets/9d8727fa-856b-4026-a5d8-55dc3bc9db99)\n![Image](https://github.com/user-attachments/assets/b2ff8e71-61c9-4588-818b-b267eb251a59)\n![Image](https://github.com/user-attachments/assets/b5ecc6e2-642e-424f-95e7-86b6bebd5366)\n![Image](https://github.com/user-attachments/assets/7d0ee718-23b9-4319-aff9-2f18a0232aff)\n![Image](https://github.com/user-attachments/assets/6d0bdf7e-cd54-40ab-8965-f55598ec4c45)\n![Image](https://github.com/user-attachments/assets/c0f42cd7-fe91-4eda-9881-adfc4dda01b9)\n\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/301/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/300", "id": 3745740435, "node_id": "I_kwDOPT-eoM7fQ3aT", "number": 300, "title": "wan animate replace mode preprocess error", "user": { "login": "rex-29", "id": 56907042, "node_id": "MDQ6VXNlcjU2OTA3MDQy", "avatar_url": "https://avatars.githubusercontent.com/u/56907042?v=4", "gravatar_id": "", "url": "https://api.github.com/users/rex-29", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-19T02:24:32Z", "updated_at": "2025-12-19T02:24:32Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi, i just meet this error as following:\n```\n2025-12-19 10:14:01,392 - INFO - [PID:574672] - propagate in video: 99%|█████████▊| 135/137 [00:27<00:00, 4.83it/s]\n2025-12-19 10:14:01,393 - INFO - [PID:574672] - propagate in video: 99%|█████████▉| 136/137 [00:27<00:00, 4.82it/s]\n2025-12-19 10:14:01,394 - INFO - [PID:574672] - propagate in video: 100%|██████████| 137/137 [00:27<00:00, 4.89it/s]\n2025-12-19 10:14:03,820 - INFO - [PID:574672] - Traceback (most recent call last):\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - File \"Wan2.2/./wan/modules/animate/preprocess/preprocess_data.py\", line 115, in \n2025-12-19 10:14:03,821 - INFO - [PID:574672] - process_pipeline(video_path=args.video_path,\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - File \"Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 111, in __call__\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - each_aug_mask = get_aug_mask(each_mask, w_len=w_len, h_len=h_len)\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - File \"Wan2.2/wan/modules/animate/preprocess/utils.py\", line 26, in get_aug_mask\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - body_bbox = get_mask_boxes(body_mask)\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - File \"Wan2.2/wan/modules/animate/preprocess/utils.py\", line 17, in get_mask_boxes\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - x_min = x_coords.min()\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - File \"uv_envs/wan/lib/python3.10/site-packages/numpy/core/_methods.py\", line 45, in _amin\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - return umr_minimum(a, axis, None, out, keepdims, initial, where)\n2025-12-19 10:14:03,821 - INFO - [PID:574672] - ValueError: zero-size array to reduction operation minimum which has no identity\n2025-12-19 10:14:07,297 - ERROR - [PID:574672] - ❌ 任务失败 (返回码 1)\n2025-12-19 10:14:07,297 - ERROR - [PID:574672] - ❌ [1/5] 失败: 一批_image_004_主体_1_video_004_视频\n2025-12-19 10:14:07,298 - INFO - [PID:574672] - 📌 [2/5] 处理中... 图片: 005_主体_4, 视频: 005_视频\n```\nI never seen it before, it works in other videos and images, is that model did not get the mask? Thanks!", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/300/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/298", "id": 3737247890, "node_id": "I_kwDOPT-eoM7eweCS", "number": 298, "title": "What is the minimum requirement for GPU to deploy this?", "user": { "login": "goodpeter-sun", "id": 107079252, "node_id": "U_kgDOBmHmVA", "avatar_url": "https://avatars.githubusercontent.com/u/107079252?v=4", "gravatar_id": "", "url": "https://api.github.com/users/goodpeter-sun", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-17T04:07:26Z", "updated_at": "2025-12-17T04:07:26Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/298/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/297", "id": 3733425121, "node_id": "I_kwDOPT-eoM7eh4vh", "number": 297, "title": "如何解决文字失真问题呢 ？", "user": { "login": "ppbcn", "id": 48338206, "node_id": "MDQ6VXNlcjQ4MzM4MjA2", "avatar_url": "https://avatars.githubusercontent.com/u/48338206?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ppbcn", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-16T06:41:53Z", "updated_at": "2025-12-16T06:41:53Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "我在图生视频的场景中，经常会遇到文字变形，sku变形等问题，请问有没有什么办法能通过后训练来解决这个问题？ ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/297/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/295", "id": 3718721958, "node_id": "I_kwDOPT-eoM7dpzGm", "number": 295, "title": "单卡4090报错CUDA out of memory，添加了--t5_cpu也没用", "user": { "login": "JohnZhuYX", "id": 7245669, "node_id": "MDQ6VXNlcjcyNDU2Njk=", "avatar_url": "https://avatars.githubusercontent.com/u/7245669?v=4", "gravatar_id": "", "url": "https://api.github.com/users/JohnZhuYX", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-12-11T10:15:20Z", "updated_at": "2025-12-15T08:59:02Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "这是我的命令python generate.py --task ti2v-5B --size 1280*704 --ckpt_dir /home/2tdisk/working/AI/model/Wan2.2-TI2V-5B --offload_model True --convert_model_dtype --t5_cpu --frame_num 9 --prompt \"Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage\"\n\n\n这时报错信息：\n\n(/home/2tdisk/env/anaconda-env/python3.10) johnzyx@johnzyx-System-Product-Name:~/working/pythonprojects/Wan2.2$ python generate.py --task ti2v-5B --size 1280*704 --ckpt_dir /home/2tdisk/working/AI/model/Wan2.2-TI2V-5B --offload_model True --convert_model_dtype --t5_cpu --frame_num 9 --prompt \"Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage\"\n[2025-12-11 17:49:00,550] INFO: Generation job args: Namespace(task='ti2v-5B', size='1280*704', frame_num=9, ckpt_dir='/home/2tdisk/working/AI/model/Wan2.2-TI2V-5B', offload_model=True, ulysses_size=1, t5_fsdp=False, t5_cpu=True, dit_fsdp=False, save_file=None, prompt='Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage', use_prompt_extend=False, prompt_extend_method='local_qwen', prompt_extend_model=None, prompt_extend_target_lang='zh', base_seed=2646853457180984931, image=None, sample_solver='unipc', sample_steps=50, sample_shift=5.0, sample_guide_scale=5.0, convert_model_dtype=True, src_root_path=None, refert_num=77, replace_flag=False, use_relighting_lora=False, num_clip=None, audio=None, enable_tts=False, tts_prompt_audio=None, tts_prompt_text=None, tts_text=None, pose_video=None, start_from_ref=False, infer_frames=80)\n[2025-12-11 17:49:00,550] INFO: Generation model config: {'__name__': 'Config: Wan TI2V 5B', 't5_model': 'umt5_xxl', 't5_dtype': torch.bfloat16, 'text_len': 512, 'param_dtype': torch.bfloat16, 'num_train_timesteps': 1000, 'sample_fps': 24, 'sample_neg_prompt': '色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走', 'frame_num': 121, 't5_checkpoint': 'models_t5_umt5-xxl-enc-bf16.pth', 't5_tokenizer': 'google/pth', 'vae_stride': (4, 16, 16), 'patch_size': (1, 2, 2), 'dim': 3072, 'ffn_dim': 14336, 'freq_dim': 256, 'num_heads': 24, 'num_layers': 30, 'window_size': (-1, -1), 'qk_norm': True, 'cross_attn_norm': True, 'eps': 1e-06, 'sample_shift': 5.0, 'sample_steps': 50, 'sample_guide_scale': 5.0}\n[2025-12-11 17:49:00,550] INFO: Input prompt: Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage\n[2025-12-11 17:49:00,550] INFO: Creating WanTI2V pipeline.\n[2025-12-11 17:49:30,888] INFO: loading /home/2tdisk/working/AI/model/Wan2.2-TI2V-5B/models_t5_umt5-xxl-enc-bf16.pth\n[2025-12-11 17:49:35,617] INFO: loading /home/2tdisk/working/AI/model/Wan2.2-TI2V-5B/Wan2.2_VAE.pth\n[2025-12-11 17:49:36,768] INFO: Creating WanModel from /home/2tdisk/working/AI/model/Wan2.2-TI2V-5B\nLoading checkpoint shards: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 129.09it/s]\n[2025-12-11 17:49:38,336] INFO: Generating video ...\n100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 50/50 [00:30<00:00, 1.66it/s]\nTraceback (most recent call last):\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/generate.py\", line 575, in \n generate(args)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/generate.py\", line 443, in generate\n video = wan_ti2v.generate(\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/textimage2video.py\", line 227, in generate\n return self.t2v(\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/textimage2video.py\", line 401, in t2v\n videos = self.vae.decode(x0)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 1043, in decode\n return [\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 1044, in \n self.model.decode(u.unsqueeze(0),\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 831, in decode\n out_ = self.decoder(\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1751, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1762, in _call_impl\n return forward_call(*args, **kwargs)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 700, in forward\n x = layer(x, feat_cache, feat_idx, first_chunk)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1751, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1762, in _call_impl\n return forward_call(*args, **kwargs)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 492, in forward\n x_main = module(x_main, feat_cache, feat_idx)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1751, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1762, in _call_impl\n return forward_call(*args, **kwargs)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 230, in forward\n x = layer(x, feat_cache[idx])\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1751, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/modules/module.py\", line 1762, in _call_impl\n return forward_call(*args, **kwargs)\n File \"/home/johnzyx/working/pythonprojects/Wan2.2/wan/modules/vae2_2.py\", line 40, in forward\n x = F.pad(x, padding)\n File \"/home/2tdisk/env/anaconda-env/python3.10/lib/python3.10/site-packages/torch/nn/functional.py\", line 5209, in pad\n return torch._C._nn.pad(input, pad, mode, value)\ntorch.OutOfMemoryError: CUDA out of memory. Tried to allocate 2.60 GiB. GPU 0 has a total capacity of 22.02 GiB of which 2.04 GiB is free. Including non-PyTorch memory, this process has 19.32 GiB memory in use. Of the allocated memory 15.98 GiB is allocated by PyTorch, and 2.85 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/295/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/294", "id": 3717341678, "node_id": "I_kwDOPT-eoM7dkiHu", "number": 294, "title": "Wan2.2 14B Animate 生成的视频，视频打开全是黑色的背景，啥都没有，求大神指点", "user": { "login": "houshubin", "id": 19849165, "node_id": "MDQ6VXNlcjE5ODQ5MTY1", "avatar_url": "https://avatars.githubusercontent.com/u/19849165?v=4", "gravatar_id": "", "url": "https://api.github.com/users/houshubin", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-11T01:27:25Z", "updated_at": "2025-12-11T01:27:25Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/294/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/293", "id": 3704438854, "node_id": "I_kwDOPT-eoM7czUBG", "number": 293, "title": "4x3090也报torch.OutOfMemoryError: CUDA out of memory", "user": { "login": "Czhazha", "id": 11772718, "node_id": "MDQ6VXNlcjExNzcyNzE4", "avatar_url": "https://avatars.githubusercontent.com/u/11772718?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Czhazha", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-08T03:01:18Z", "updated_at": "2025-12-08T03:01:54Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "task='i2v-A14B'\n4x3090也报torch.OutOfMemoryError: CUDA out of memory\n有什么可以调整参数的地方能降一降显存需求吗？", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/293/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/292", "id": 3688565611, "node_id": "I_kwDOPT-eoM7b2wtr", "number": 292, "title": "二次方程等式有误", "user": { "login": "jikhunb", "id": 128955174, "node_id": "U_kgDOB6-zJg", "avatar_url": "https://avatars.githubusercontent.com/u/128955174?v=4", "gravatar_id": "", "url": "https://api.github.com/users/jikhunb", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-03T05:24:27Z", "updated_at": "2025-12-03T05:24:27Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "wan/speech2video.py文件夹下的函数 def get_size_less_than_area体中，\n\nmin_scale = (-b + math.sqrt(b**2 - 2 * a * c)) / (\n 2 * a)\n\n# Calculate scale boundaries using quadratic equation\nmin_scale = (-b + math.sqrt(b**2 - 2 * a * c)) / (2 * a) # Scale when maximum padding is applied\n\n应该是：\nmin_scale = (-b + math.sqrt(b**2 - 4* a * c)) / (2 * a) # Scale when maximum padding is applied\n\n 2 * a * c --> 4* a * c\n\n以上错误由网页端qwen千问发现，大模型为qwen3-max", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/292/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/291", "id": 3685239708, "node_id": "I_kwDOPT-eoM7bqEuc", "number": 291, "title": "Wan 2.5 weights? Will be open-sourced?", "user": { "login": "Tortoise17", "id": 36593708, "node_id": "MDQ6VXNlcjM2NTkzNzA4", "avatar_url": "https://avatars.githubusercontent.com/u/36593708?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Tortoise17", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2025-12-02T10:45:30Z", "updated_at": "2025-12-17T06:04:47Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Dear Respected Support. \n\nThank you so much for a great input. I want to ask that is there any plan to release the weights of wan 2.5? Or they will not be open-sourced?\n\nPlease guide .. ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/291/reactions", "total_count": 9, "+1": 9, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/290", "id": 3684891792, "node_id": "I_kwDOPT-eoM7bovyQ", "number": 290, "title": "How to convert diffusers model to wan2.2 format", "user": { "login": "wikiwen", "id": 38649098, "node_id": "MDQ6VXNlcjM4NjQ5MDk4", "avatar_url": "https://avatars.githubusercontent.com/u/38649098?v=4", "gravatar_id": "", "url": "https://api.github.com/users/wikiwen", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-02T09:17:31Z", "updated_at": "2025-12-02T09:17:31Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Great work!\n\nI see `convert_wan_to_diffusers.py` in diffusers repo，but no `convert_diffusers_to_wan.py`. Do you have plan to upload a offical convert scripts?", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/290/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/289", "id": 3684397669, "node_id": "I_kwDOPT-eoM7bm3Jl", "number": 289, "title": "求教混合精度训练和推理", "user": { "login": "JohnHerry", "id": 8011802, "node_id": "MDQ6VXNlcjgwMTE4MDI=", "avatar_url": "https://avatars.githubusercontent.com/u/8011802?v=4", "gravatar_id": "", "url": "https://api.github.com/users/JohnHerry", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-12-02T07:03:16Z", "updated_at": "2025-12-02T07:03:16Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "感谢开源，想要学习一下，DiT based model，如何保持训练和推理时 混合精度推理的效果呢？ 之前尝试CosyVoice3 的DiT-flow-matching ，其训练时可以bf16 混合训练，但是推理之前将模型转换为 Tensorrt fp16，就会数值溢出，生成Nan。 看到Wan2.2 项目中有使用T5模块做DiT 建模，其中设计了 fp16 clamp。请问这个设计的话对模型训练的最终效果有多大负面影响吗？ 另外进行TensorRT 推理时是否可以保持效果？谢谢！", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/289/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/287", "id": 3680882692, "node_id": "I_kwDOPT-eoM7bZdAE", "number": 287, "title": "Wan-Animate预处理模型地址", "user": { "login": "gluttony-10", "id": 52977964, "node_id": "MDQ6VXNlcjUyOTc3OTY0", "avatar_url": "https://avatars.githubusercontent.com/u/52977964?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gluttony-10", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-12-01T12:53:49Z", "updated_at": "2025-12-05T11:47:02Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "麻烦给出Wan-Animate预处理模型地址\n```\n/path/to/your/ckpt_path/\n ├── det/\n │ └── yolov10m.onnx\n ├── pose2d/\n │ └── vitpose_h_wholebody.onnx\n ├── sam2/\n │ └── sam2_hiera_large.pt\n```", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/287/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/286", "id": 3662082025, "node_id": "I_kwDOPT-eoM7aRu_p", "number": 286, "title": "music", "user": { "login": "mosiartermusic", "id": 246042702, "node_id": "U_kgDODqpQTg", "avatar_url": "https://avatars.githubusercontent.com/u/246042702?v=4", "gravatar_id": "", "url": "https://api.github.com/users/mosiartermusic", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-25T08:53:49Z", "updated_at": "2025-11-25T08:53:49Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/286/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/285", "id": 3658214992, "node_id": "I_kwDOPT-eoM7aC-5Q", "number": 285, "title": "为什么运行wan2.2的视频生成工作流执行到13%就中断，文件位置都放对啦，可以运行，但运行wan2.2的视频就会中断comfyui的链接，显存也没爆", "user": { "login": "yzg19272194217-ops", "id": 226065652, "node_id": "U_kgDODXl89A", "avatar_url": "https://avatars.githubusercontent.com/u/226065652?v=4", "gravatar_id": "", "url": "https://api.github.com/users/yzg19272194217-ops", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-24T10:35:47Z", "updated_at": "2025-11-24T10:35:47Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "\"Image\"\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/285/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/284", "id": 3650818268, "node_id": "I_kwDOPT-eoM7ZmxDc", "number": 284, "title": "wan2.2 Computational Efficiency on Different GPUs", "user": { "login": "hogan6666", "id": 218532588, "node_id": "U_kgDODQaK7A", "avatar_url": "https://avatars.githubusercontent.com/u/218532588?v=4", "gravatar_id": "", "url": "https://api.github.com/users/hogan6666", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-21T09:05:19Z", "updated_at": "2025-11-21T09:05:19Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "https://github.com/Wan-Video/Wan2.2\n\nFor T2V case, we know that the inference time is 2735.7s at A100. What's the frame number? 81frames ?\n\nThanks.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/284/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/283", "id": 3650060160, "node_id": "I_kwDOPT-eoM7Zj3-A", "number": 283, "title": "OutOfMemoryError: CUDA out of memory Animate-14B", "user": { "login": "rbur0425", "id": 14355895, "node_id": "MDQ6VXNlcjE0MzU1ODk1", "avatar_url": "https://avatars.githubusercontent.com/u/14355895?v=4", "gravatar_id": "", "url": "https://api.github.com/users/rbur0425", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-11-21T03:33:37Z", "updated_at": "2025-11-21T17:25:04Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "[2025-11-20 21:19:21,501] INFO: Generation job args: Namespace(task='animate-14B', size='1280*720', frame_num=77, ckpt_dir='./Wan2.2-Animate-14B', offload_model=True, ulysses_size=1, t5_fsdp=False, t5_cpu=True, dit_fsdp=False, save_file=None, prompt='视频中的人在做动作', use_prompt_extend=False, prompt_extend_method='local_qwen', prompt_extend_model=None, prompt_extend_target_lang='zh', base_seed=3912865311457787091, image=None, sample_solver='unipc', sample_steps=20, sample_shift=5.0, sample_guide_scale=1.0, convert_model_dtype=True, src_root_path='./replacement_output/process_results', refert_num=1, replace_flag=True, use_relighting_lora=True, num_clip=None, audio=None, enable_tts=False, tts_prompt_audio=None, tts_prompt_text=None, tts_text=None, pose_video=None, start_from_ref=False, infer_frames=80)\n[2025-11-20 21:19:21,501] INFO: Generation model config: {'__name__': 'Config: Wan animate 14B', 't5_model': 'umt5_xxl', 't5_dtype': torch.bfloat16, 'text_len': 512, 'param_dtype': torch.bfloat16, 'num_train_timesteps': 1000, 'sample_fps': 30, 'sample_neg_prompt': '色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走', 'frame_num': 77, 't5_checkpoint': 'models_t5_umt5-xxl-enc-bf16.pth', 't5_tokenizer': 'google/umt5-xxl', 'clip_checkpoint': 'models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth', 'clip_tokenizer': 'xlm-roberta-large', 'lora_checkpoint': 'relighting_lora.ckpt', 'vae_checkpoint': 'Wan2.1_VAE.pth', 'vae_stride': (4, 8, 8), 'patch_size': (1, 2, 2), 'dim': 5120, 'ffn_dim': 13824, 'freq_dim': 256, 'num_heads': 40, 'num_layers': 40, 'window_size': (-1, -1), 'qk_norm': True, 'cross_attn_norm': True, 'eps': 1e-06, 'use_face_encoder': True, 'motion_encoder_dim': 512, 'sample_shift': 5.0, 'sample_steps': 20, 'sample_guide_scale': 1.0, 'prompt': '视频中的人在做动作'}\n[2025-11-20 21:19:21,502] INFO: Input prompt: 视频中的人在做动作\n[2025-11-20 21:19:21,502] INFO: Creating Wan-Animate pipeline.\n[2025-11-20 21:20:37,088] INFO: loading ./Wan2.2-Animate-14B/models_t5_umt5-xxl-enc-bf16.pth\n[2025-11-20 21:20:45,665] INFO: loading ./Wan2.2-Animate-14B/models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth\n[2025-11-20 21:20:52,849] INFO: loading ./Wan2.2-Animate-14B/Wan2.1_VAE.pth\n[2025-11-20 21:20:53,660] INFO: Creating WanAnimate from ./Wan2.2-Animate-14B\nLoading checkpoint shards: 100%|████████████████████████████████████████████████████████| 4/4 [00:43<00:00, 10.76s/it]\n[2025-11-20 21:21:37,671] INFO: Loading Relighting Lora.\n[2025-11-20 21:21:51,071] INFO: Generating video ...\n[2025-11-20 22:19:53,554] INFO: real frames: 1142 target frames: 1217\nTraceback (most recent call last):\n File \"/home/roland/Sites/Wan2.2/generate.py\", line 575, in \n generate(args)\n File \"/home/roland/Sites/Wan2.2/generate.py\", line 471, in generate\n video = wan_animate.generate(\n ^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/animate.py\", line 564, in generate\n chunk_latents = self.vae.encode([chunk.to(torch.bfloat16)])[0]\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/modules/vae2_1.py\", line 653, in encode\n self.model.encode(u.unsqueeze(0), self.scale).float().squeeze(0)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/modules/vae2_1.py\", line 530, in encode\n out_ = self.encoder(\n ^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1736, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1747, in _call_impl\n return forward_call(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/modules/vae2_1.py\", line 338, in forward\n x = layer(x, feat_cache, feat_idx)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1736, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1747, in _call_impl\n return forward_call(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/modules/vae2_1.py\", line 215, in forward\n x = layer(x, feat_cache[idx])\n ^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1736, in _wrapped_call_impl\n return self._call_impl(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/modules/module.py\", line 1747, in _call_impl\n return forward_call(*args, **kwargs)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/wan/modules/vae2_1.py\", line 34, in forward\n x = F.pad(x, padding)\n ^^^^^^^^^^^^^^^^^\n File \"/home/roland/Sites/Wan2.2/.venv/lib/python3.12/site-packages/torch/nn/functional.py\", line 5096, in pad\n return torch._C._nn.pad(input, pad, mode, value)\n ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^\ntorch.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.99 GiB. GPU 0 has a total capacity of 47.54 GiB of which 1.79 GiB is free. Including non-PyTorch memory, this process has 45.73 GiB memory in use. Of the allocated memory 45.32 GiB is allocated by PyTorch, and 103.44 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)\n\n\nIs it possible to run this on a A6000 GPU 48 GB VRAM? \n\npython 3.12.3\ntorch 2.5.1+cu121 (CUDA 12.1, CUDA available True)\nflash-attn 2.8.3\ntransformers 4.51.3\npeft 0.17.0\nonnxruntime-gpu 1.23.2\nGPU: NVIDIA RTX A6000 (driver 535.274.02, ~48 GB)", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/283/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/282", "id": 3646344079, "node_id": "I_kwDOPT-eoM7ZVsuP", "number": 282, "title": "May i ask you how much time it costs to generate video?", "user": { "login": "YiCheng-W992", "id": 177719760, "node_id": "U_kgDOCpfJ0A", "avatar_url": "https://avatars.githubusercontent.com/u/177719760?v=4", "gravatar_id": "", "url": "https://api.github.com/users/YiCheng-W992", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-11-20T09:31:40Z", "updated_at": "2025-12-06T07:57:56Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I use single 4090 to generate \n\npython generate.py --task ti2v-5B --size 1280*704 --ckpt_dir ../Models/Wan2.2-TI2V-5B --offload_model True --convert_model_dtype --t5_cpu --prompt \"Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage\" \n\n\n[2025-11-20 16:41:29,023] INFO: Input prompt: Two anthropomorphic cats in comfy boxing gear and bright gloves fight intensely on a spotlighted stage\n[2025-11-20 16:41:29,023] INFO: Creating WanTI2V pipeline.\n[2025-11-20 16:42:41,146] INFO: loading ../Models/Wan2.2-TI2V-5B/models_t5_umt5-xxl-enc-bf16.pth\n[2025-11-20 16:43:27,652] INFO: loading ../Models/Wan2.2-TI2V-5B/Wan2.2_VAE.pth\n[2025-11-20 16:43:38,051] INFO: Creating WanModel from ../Models/Wan2.2-TI2V-5B\nLoading checkpoint shards: 100%|███████████████████████████████████████████████████████████████████████████████████| 3/3 [00:00<00:00, 4.00it/s]\n[2025-11-20 16:44:49,164] INFO: Generating video ...\n\n\nHowever, an hour has been used ,there is no other message , is it normal ?\n\nGPU util is zero and memory usage is 3GB", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/282/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/281", "id": 3635925958, "node_id": "I_kwDOPT-eoM7Yt9PG", "number": 281, "title": "微调训练框架选择", "user": { "login": "YueMing-github", "id": 67255039, "node_id": "MDQ6VXNlcjY3MjU1MDM5", "avatar_url": "https://avatars.githubusercontent.com/u/67255039?v=4", "gravatar_id": "", "url": "https://api.github.com/users/YueMing-github", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-11-18T03:19:21Z", "updated_at": "2025-11-21T11:09:13Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "请问如果想微调wan2.2-animate-move 有没有训练框架推荐 感谢 ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/281/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/280", "id": 3632469087, "node_id": "I_kwDOPT-eoM7YgxRf", "number": 280, "title": "what is the function of adding clip_context?", "user": { "login": "parryppp", "id": 165291437, "node_id": "U_kgDOCdolrQ", "avatar_url": "https://avatars.githubusercontent.com/u/165291437?v=4", "gravatar_id": "", "url": "https://api.github.com/users/parryppp", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-17T09:26:06Z", "updated_at": "2025-11-17T09:27:08Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "https://github.com/Wan-Video/Wan2.2/blob/main/wan/animate.py#L586\n\n\"Image\"\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/280/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/279", "id": 3629547968, "node_id": "I_kwDOPT-eoM7YVoHA", "number": 279, "title": "请问amd7900显卡可以运行Wan2.2吗", "user": { "login": "963431755", "id": 7754091, "node_id": "MDQ6VXNlcjc3NTQwOTE=", "avatar_url": "https://avatars.githubusercontent.com/u/7754091?v=4", "gravatar_id": "", "url": "https://api.github.com/users/963431755", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-16T00:55:51Z", "updated_at": "2025-11-16T00:55:51Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/279/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/278", "id": 3629350593, "node_id": "I_kwDOPT-eoM7YU37B", "number": 278, "title": "Libraries not mentioned in requirements.txt", "user": { "login": "AshishRaj04", "id": 122210726, "node_id": "U_kgDOB0jJpg", "avatar_url": "https://avatars.githubusercontent.com/u/122210726?v=4", "gravatar_id": "", "url": "https://api.github.com/users/AshishRaj04", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2025-11-15T22:26:54Z", "updated_at": "2025-11-20T09:48:28Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "The tools and libraries that are used by the model are not mentioned in the requirements.txt file .\nHence while inferencing the model I getting errors like\n`\nFile \"/workspace/Wan2.2/wan/modules/s2v/audio_encoder.py\", line 4, in \n import librosa\nModuleNotFoundError: No module named 'librosa'\n`\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/278/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/277", "id": 3628801540, "node_id": "I_kwDOPT-eoM7YSx4E", "number": 277, "title": "क्या तुम भी किसी का दिमाग पढ़ना चाहते हो?तो तुम्हें सिर्फ यह तीन बातों का ध्यान रखना चाहिए", "user": { "login": "vy124386-cyber", "id": 244195689, "node_id": "U_kgDODo4haQ", "avatar_url": "https://avatars.githubusercontent.com/u/244195689?v=4", "gravatar_id": "", "url": "https://api.github.com/users/vy124386-cyber", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2025-11-15T14:48:40Z", "updated_at": "2025-11-15T22:24:17Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "#111 ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/277/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/273", "id": 3600056325, "node_id": "I_kwDOPT-eoM7WlIAF", "number": 273, "title": "wan 2.2 cause nan in latents", "user": { "login": "chaowenguo", "id": 229142208, "node_id": "U_kgDODahuwA", "avatar_url": "https://avatars.githubusercontent.com/u/229142208?v=4", "gravatar_id": "", "url": "https://api.github.com/users/chaowenguo", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-11-07T12:08:27Z", "updated_at": "2025-11-25T09:35:40Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "here is the source code:\nimport asyncio, aiohttp.web, math, pathlib, torch, builtins, transformers, diffusers, av, io, numpy, uvloop, concurrent, multiprocessing, os, zipfile, sys\nfrom modelscope.hub.snapshot_download import snapshot_download\nfrom modelscope.hub.file_download import model_file_download\n#import sd_embed.embedding_funcs\n\nmodel_id = 'Wan-AI/Wan2.2-T2V-A14B-Diffusers'\nmodel_dir = snapshot_download(model_id)\nonload_device = torch.device('cuda')\npipe = diffusers.WanPipeline.from_pretrained(model_dir, vae=diffusers.AutoencoderKLWan.from_pretrained(model_dir, subfolder='vae', torch_dtype=torch.float32).to(onload_device),\ndtype=torch.bfloat16, low_cpu_mem_usage=True,\nquantization_config=diffusers.PipelineQuantizationConfig(quant_backend='bitsandbytes_4bit', quant_kwargs={'load_in_4bit':True, 'bnb_4bit_quant_type':'nf4', 'bnb_4bit_compute_dtype':torch.bfloat16}, components_to_quantize=['transformer', 'transformer_2']))\ndiffusers.hooks.apply_group_offloading(pipe.text_encoder, onload_device=onload_device, offload_type='leaf_level', use_stream=True, record_stream=True, non_blocking=True, low_cpu_mem_usage=True)\npipe.transformer.enable_group_offload(onload_device=onload_device, offload_type='leaf_level', use_stream=True, record_stream=True, non_blocking=True, low_cpu_mem_usage=True)\npipe.transformer.enable_layerwise_casting(storage_dtype=torch.float8_e4m3fn, compute_dtype=torch.bfloat16)\npipe.transformer_2.enable_group_offload(onload_device=onload_device, offload_type='leaf_level', use_stream=True, record_stream=True, non_blocking=True, low_cpu_mem_usage=True)\npipe.transformer_2.enable_layerwise_casting(storage_dtype=torch.float8_e4m3fn, compute_dtype=torch.bfloat16)\npipe.vae.enable_tiling()\npipe.vae.enable_slicing()\n\ndef decode_tensors(pipe, step, timestep, callback_kwargs):\nlatents = callback_kwargs[\"latents\"]\nprint(step, latents)\nreturn callback_kwargs\n\nwith zipfile.ZipFile('output.zip', 'w') as output:\nbuffer = io.BytesIO()\nwith av.open(buffer, mode='w', format='mp4') as writer:\nstream = writer.add_stream('h264', rate=8)\nstream.width = 1280\nstream.height = 720\nstream.pix_fmt = 'yuv420p'\n#prompt_embeds, prompt_neg_embeds = sd_embed.embedding_funcs.get_weighted_text_embeddings_sd15(pipe, prompt=prompt, neg_prompt=negative)\nfor frame in pipe(prompt=prompt, negative_prompt=negative, width=stream.width, height=stream.height, num_frames=81, num_inference_steps=5, callback_on_step_end=decode_tensors).frames[0]: writer.mux(stream.encode(av.VideoFrame.from_ndarray((frame * 255).astype(numpy.uint8))))\nwriter.mux(stream.encode())\noutput.writestr('output.mp4', buffer.getvalue())\n\nprint(step, latents) all the latents are nan please help to solve the problem.\n\nubuntu22.04-cuda12.4.0-py311-torch2.8.0, diffusers 0.35.2", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/273/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/272", "id": 3599223686, "node_id": "I_kwDOPT-eoM7Wh8uG", "number": 272, "title": "Wan 2.2 微调：noisy reference latent对应的loss如何计算？求助help", "user": { "login": "0xLDF", "id": 64844447, "node_id": "MDQ6VXNlcjY0ODQ0NDQ3", "avatar_url": "https://avatars.githubusercontent.com/u/64844447?v=4", "gravatar_id": "", "url": "https://api.github.com/users/0xLDF", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-07T08:18:08Z", "updated_at": "2025-11-07T08:18:08Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi Wan team,\nI’m trying to fine-tune Wan-2.2-Animate and would like to clarify how the loss is calculated for the two “extra” latents:\n```\nnoisy reference latents\nnoisy temporal latents\n```\nFrom the inference code I see that the denoising network receives three kinds of noisy latents:\n```\nnoisy video latents\nnoisy reference latents\nnoisy temporal latents\n```\nThe reference latents is also denoised step-by-step, so their value changes during sampling.\nThis implies that you must have computed a loss on both the reference and the temporal latent during training. Could you share the details?\nDid you simply apply the same flow-matching loss to all three tensors and sum/average them?\n\nFinally, I would like to ask: in current image/video generation models such as Flux-Kontext, Stand-in, Phantom, and RealisDance-DiT, the noisy latents never contain the reference image, and reference is not supervised by any loss during training. What was the motivation behind Wan-2.2-Animate’s decision to add noise to the reference latents.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/272/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/271", "id": 3594844677, "node_id": "I_kwDOPT-eoM7WRPoF", "number": 271, "title": "wan-s2v drop_first_motion的作用", "user": { "login": "liwang0621", "id": 81345643, "node_id": "MDQ6VXNlcjgxMzQ1NjQz", "avatar_url": "https://avatars.githubusercontent.com/u/81345643?v=4", "gravatar_id": "", "url": "https://api.github.com/users/liwang0621", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-06T09:11:28Z", "updated_at": "2025-11-06T09:11:28Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "看代码的时候有个疑问，当drop_first_motion=True and r==0的时候，为什么要去掉前三帧呀？\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/271/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/270", "id": 3594392673, "node_id": "I_kwDOPT-eoM7WPhRh", "number": 270, "title": "wan2.2各渠道的版本及分辨率问题", "user": { "login": "lzliii", "id": 108640450, "node_id": "U_kgDOBnm4wg", "avatar_url": "https://avatars.githubusercontent.com/u/108640450?v=4", "gravatar_id": "", "url": "https://api.github.com/users/lzliii", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-11-06T06:57:46Z", "updated_at": "2025-11-06T06:57:46Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "在计划使用wan2.2模型推理时，发现各个来源渠道的支持内容都不一样，进而不清楚该选用哪种能力。有大佬能帮忙看看，确认下背后的逻辑吗？以及各个模型的关联\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/270/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/269", "id": 3575723277, "node_id": "I_kwDOPT-eoM7VITUN", "number": 269, "title": "CUDA OOM for 193frames of I2V using 4 H100. 图生视频 193 帧，4 张 H100爆显存", "user": { "login": "lxrswdd", "id": 35613337, "node_id": "MDQ6VXNlcjM1NjEzMzM3", "avatar_url": "https://avatars.githubusercontent.com/u/35613337?v=4", "gravatar_id": "", "url": "https://api.github.com/users/lxrswdd", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-31T16:59:31Z", "updated_at": "2025-10-31T17:00:45Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "为什么4 个h100-80gb跑 193 帧的视频都会爆显存？\n\nCUDA OOM for 193frames of I2V using 4 H100.\n14B 模型，设置如下。\n```\ntorchrun --nproc_per_node=4 generate.py --task i2v-A14B --size 720*1280 \\\n --ckpt_dir ./models/Wan2.2-I2V-A14B --offload_model True --convert_model_dtype \\\n --image /home/ubuntu/data/image/man_spin-0002.png \\\n --prompt \"A man sits on a chair and holds a camera in his hand. He is spinning on this turnable chair.\" \\\n --frame_num 193 \\\n --offload_model True --convert_model_dtype \\\n --dit_fsdp --t5_fsdp --ulysses_size 4 \\\n``` \n\n刚启动就如下的显存占用，然后跑一半显示爆显存。\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/269/reactions", "total_count": 2, "+1": 2, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/266", "id": 3570540479, "node_id": "I_kwDOPT-eoM7U0h-_", "number": 266, "title": "Why do we need tpl_pose_metas_new list in get_retarget_pose function", "user": { "login": "abdelrahman-shaaban98", "id": 52332496, "node_id": "MDQ6VXNlcjUyMzMyNDk2", "avatar_url": "https://avatars.githubusercontent.com/u/52332496?v=4", "gravatar_id": "", "url": "https://api.github.com/users/abdelrahman-shaaban98", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-30T12:24:37Z", "updated_at": "2025-10-30T12:24:37Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Inside the get_retarget_pose function in retarget_pose.py, why do we need tpl_pose_metas_new list, isn't it the same list as tpl_pose_metas?", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/266/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/264", "id": 3564995849, "node_id": "I_kwDOPT-eoM7UfYUJ", "number": 264, "title": "请问是否有支持微调wan2.2 animate的项目？最好能单卡跑", "user": { "login": "professorhuanghaojing", "id": 178784988, "node_id": "U_kgDOCqgK3A", "avatar_url": "https://avatars.githubusercontent.com/u/178784988?v=4", "gravatar_id": "", "url": "https://api.github.com/users/professorhuanghaojing", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-29T09:04:21Z", "updated_at": "2025-10-29T09:04:21Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/264/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/262", "id": 3561587301, "node_id": "I_kwDOPT-eoM7USYJl", "number": 262, "title": "Can anyone tell me what are the adequate/ ideal hardware requirements for the Wan2.2-animate animation model?", "user": { "login": "digilateraldev", "id": 193566282, "node_id": "U_kgDOC4mWSg", "avatar_url": "https://avatars.githubusercontent.com/u/193566282?v=4", "gravatar_id": "", "url": "https://api.github.com/users/digilateraldev", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-28T13:29:42Z", "updated_at": "2025-10-31T23:10:11Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I am struggling to get an idea if the model will perfectly run on which gpu with what configuration.\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/262/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/257", "id": 3554637911, "node_id": "I_kwDOPT-eoM7T33hX", "number": 257, "title": "Possibility of passing latent instead of image in continuous i2v workflow", "user": { "login": "anitman", "id": 3346288, "node_id": "MDQ6VXNlcjMzNDYyODg=", "avatar_url": "https://avatars.githubusercontent.com/u/3346288?v=4", "gravatar_id": "", "url": "https://api.github.com/users/anitman", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-27T00:40:15Z", "updated_at": "2025-10-27T00:42:53Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "In our Image-to-Video (I2V) generation process, I generate 5-second videos at 16 FPS, resulting in 81 frames. I’ve noticed that when I use the pixel image of the 81st frame (x81x_{81}x_{81}) as the starting input for the next round of video generation, the output quality degrades over time (e.g., objects distort or textures blur). I believe this happens due to errors introduced when decoding and re-encoding the pixel image.To avoid this degradation, I want to extract the latent representation of the 81st frame (z81z_{81}z_{81}) directly from the latent space before it’s decoded into a pixel image and use it as the starting input for the next round of generation. This way, we can continue generating from the 81st frame without losing information.\n\nMy questions are:\n\n1.Is it technically possible to extract the latent representation (z81z_{81}z_{81}) of the 81st frame from the diffusion model’s output before decoding it to a pixel image (x81x_{81}x_{81})?\n\n2.If possible, how can we implement this in the generation pipeline?\nWill using z81z_{81}z_{81} as the input for the next round help maintain video consistency and reduce degradation compared to using the pixel image x81x_{81}x_{81}?\n\nAre there any limitations or challenges I should be aware of when extracting and reusing z81z_{81}z_{81}, such as model constraints or potential drift in long-term generation?\n\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/257/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/256", "id": 3552177121, "node_id": "I_kwDOPT-eoM7Tuevh", "number": 256, "title": "Stuck on rendering. Need to cancel the generations but cant", "user": { "login": "melanieb5277-ux", "id": 240144547, "node_id": "U_kgDODlBQow", "avatar_url": "https://avatars.githubusercontent.com/u/240144547?v=4", "gravatar_id": "", "url": "https://api.github.com/users/melanieb5277-ux", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 6, "created_at": "2025-10-25T09:11:01Z", "updated_at": "2025-11-02T17:13:27Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I am stuck on 3 renders that are simnply not worknig - but i cannot delete them.\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/256/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/255", "id": 3552168878, "node_id": "I_kwDOPT-eoM7Tucuu", "number": 255, "title": "wan animate output length limit", "user": { "login": "rubentorresbonet", "id": 4907670, "node_id": "MDQ6VXNlcjQ5MDc2NzA=", "avatar_url": "https://avatars.githubusercontent.com/u/4907670?v=4", "gravatar_id": "", "url": "https://api.github.com/users/rubentorresbonet", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-25T09:06:20Z", "updated_at": "2025-10-25T09:06:20Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi,\n\nI saw somewhere between 2-30s is supported, but I wonder if 5 minute animate for avatar replacement is feasible?\n\nHow does compute requirements scale, is it just more time or does memory also skyrocket?\n\nThanks", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/255/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/253", "id": 3543793650, "node_id": "I_kwDOPT-eoM7TOf_y", "number": 253, "title": "40G的A100跑T2V需要多少显存和内存？", "user": { "login": "fhkankan", "id": 33663480, "node_id": "MDQ6VXNlcjMzNjYzNDgw", "avatar_url": "https://avatars.githubusercontent.com/u/33663480?v=4", "gravatar_id": "", "url": "https://api.github.com/users/fhkankan", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-23T09:03:33Z", "updated_at": "2025-10-23T11:08:28Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/253/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/252", "id": 3540013199, "node_id": "I_kwDOPT-eoM7TAFCP", "number": 252, "title": "H", "user": { "login": "asmatullahazkhan-source", "id": 236727719, "node_id": "U_kgDODhwtpw", "avatar_url": "https://avatars.githubusercontent.com/u/236727719?v=4", "gravatar_id": "", "url": "https://api.github.com/users/asmatullahazkhan-source", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-22T09:36:03Z", "updated_at": "2025-10-25T03:27:55Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/252/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/251", "id": 3540005421, "node_id": "I_kwDOPT-eoM7TADIt", "number": 251, "title": "Asmat", "user": { "login": "asmatullahazkhan-source", "id": 236727719, "node_id": "U_kgDODhwtpw", "avatar_url": "https://avatars.githubusercontent.com/u/236727719?v=4", "gravatar_id": "", "url": "https://api.github.com/users/asmatullahazkhan-source", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-22T09:34:25Z", "updated_at": "2025-10-22T09:34:25Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/251/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/250", "id": 3539805327, "node_id": "I_kwDOPT-eoM7S_SSP", "number": 250, "title": "HF 和ModelScope上的demo都不能用了，麻烦看一下，谢谢", "user": { "login": "jakedong93-droid", "id": 239552942, "node_id": "U_kgDODkdJrg", "avatar_url": "https://avatars.githubusercontent.com/u/239552942?v=4", "gravatar_id": "", "url": "https://api.github.com/users/jakedong93-droid", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-22T08:54:55Z", "updated_at": "2025-10-22T08:54:55Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/250/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/248", "id": 3536127615, "node_id": "I_kwDOPT-eoM7SxQZ_", "number": 248, "title": "boboleta voando com um elelfante nas costas", "user": { "login": "netogf", "id": 154078159, "node_id": "U_kgDOCS8Lzw", "avatar_url": "https://avatars.githubusercontent.com/u/154078159?v=4", "gravatar_id": "", "url": "https://api.github.com/users/netogf", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-10-21T11:32:35Z", "updated_at": "2025-10-21T19:25:45Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "boboleta voando com um elelfante nas costas\n\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/248/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/247", "id": 3533296932, "node_id": "I_kwDOPT-eoM7SmdUk", "number": 247, "title": "Mulher loira", "user": { "login": "bandapaixaonordeste-oss", "id": 239206272, "node_id": "U_kgDODkH_gA", "avatar_url": "https://avatars.githubusercontent.com/u/239206272?v=4", "gravatar_id": "", "url": "https://api.github.com/users/bandapaixaonordeste-oss", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-20T17:28:44Z", "updated_at": "2025-10-20T17:28:44Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/247/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/246", "id": 3533296057, "node_id": "I_kwDOPT-eoM7SmdG5", "number": 246, "title": "Teozinho", "user": { "login": "bandapaixaonordeste-oss", "id": 239206272, "node_id": "U_kgDODkH_gA", "avatar_url": "https://avatars.githubusercontent.com/u/239206272?v=4", "gravatar_id": "", "url": "https://api.github.com/users/bandapaixaonordeste-oss", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-20T17:28:23Z", "updated_at": "2025-10-20T17:28:23Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Faz um vídeo de uma mulher loira com olhos cor de mel anando de moto numa tarde e verão ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/246/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/245", "id": 3532449317, "node_id": "I_kwDOPT-eoM7SjOYl", "number": 245, "title": "Request for Additional Information – Wan-Video/Wan2.2", "user": { "login": "geovanehenrique1521-png", "id": 239165666, "node_id": "U_kgDODkFg4g", "avatar_url": "https://avatars.githubusercontent.com/u/239165666?v=4", "gravatar_id": "", "url": "https://api.github.com/users/geovanehenrique1521-png", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-20T13:26:35Z", "updated_at": "2025-10-20T13:26:35Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "This is a placeholder issue for the Wan-Video/Wan2.2 repository. Please provide more details about your bug report, feature request, or task. Helpful information includes:\n- The type of issue (bug, feature request, task, etc.)\n- A detailed description of the problem or request\n- Steps to reproduce (if applicable)\n- Expected behavior\n- Relevant files, screenshots, or code snippets\n\nOnce you provide the requested information, the issue can be updated to include all necessary context.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/245/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/242", "id": 3529382104, "node_id": "I_kwDOPT-eoM7SXhjY", "number": 242, "title": "Geninho", "user": { "login": "lucasviana91-dotcom", "id": 238921052, "node_id": "U_kgDODj2lXA", "avatar_url": "https://avatars.githubusercontent.com/u/238921052?v=4", "gravatar_id": "", "url": "https://api.github.com/users/lucasviana91-dotcom", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-19T00:50:06Z", "updated_at": "2025-10-19T00:50:06Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Preserve 100% a identidade do personagem da imagem de referência:\n— Rosto: formato oval, testa alta com entradas discretas, sorriso largo com dentes à mostra, óculos retangulares pretos, orelhas discretas.\n— Pele: tom médio claro (mesmo tom da referência).\n— Corpo: proporção adulta, tronco curto, braços fortes, mãos grandes.\n— Roupa: camisa polo vermelha com gola e abotoamento escuros; logotipo Geninho em branco no peito esquerdo (mesma fonte/símbolo); calça azul-marinho; cinto marrom; bolsos/bolsas de ferramentas nas laterais; botas marrons.\n— Proibições: não alterar cor do uniforme, não adicionar capacete, barba, gorro, bigode, tatuagens, adesivos novos ou logotipos diferentes; não mudar formato dos óculos.\n\nCena: corredor de loja de materiais de construção, prateleiras desfocadas ao fundo, luz quente e comercial.\nAção: Geninho segura um incinerador/lança-promoções (prop de marketing, preto e laranja). Ao acionar, surge efeito de chama VFX controlada acima do bocal, sem queimar nada real. Etiquetas/flâmulas flutuantes com símbolo % e textos “OFERTA”, “QUEIMA DE PREÇOS”, “QUEIMA HOJE” surgem e brilham (glow).\nCâmera: plano americano com leve dolly-in; depois close no bocal e nas etiquetas; fim em plano mais aberto com o corredor.\nLook & segurança: contraste alto, bloom suave no VFX, partículas douradas; nenhum objeto pega fogo de verdade; chamas são apenas efeito digital publicitário.\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/242/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/241", "id": 3528469771, "node_id": "I_kwDOPT-eoM7SUC0L", "number": 241, "title": "Nachrichten", "user": { "login": "maikel666", "id": 211297085, "node_id": "U_kgDODJgjPQ", "avatar_url": "https://avatars.githubusercontent.com/u/211297085?v=4", "gravatar_id": "", "url": "https://api.github.com/users/maikel666", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-18T09:41:20Z", "updated_at": "2025-10-18T09:41:20Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Erstelle einen Trailer für eine Nachrichtensendung an Anfang ein Polizeiwagen in Thailand mit Blaulicht, dann die Polizisten einen Mann verfolgen und thailändische Strand, wo ein Schnellboot lang fährt und die Polizei dahinter", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/241/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/240", "id": 3525584740, "node_id": "I_kwDOPT-eoM7SJCdk", "number": 240, "title": "Porsche is driving in the mountains", "user": { "login": "kapilarprojects-dotcom", "id": 238669502, "node_id": "U_kgDODjnOvg", "avatar_url": "https://avatars.githubusercontent.com/u/238669502?v=4", "gravatar_id": "", "url": "https://api.github.com/users/kapilarprojects-dotcom", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-17T10:53:02Z", "updated_at": "2025-10-17T10:53:02Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Black Porsche \n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/240/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/239", "id": 3525028110, "node_id": "I_kwDOPT-eoM7SG6kO", "number": 239, "title": "KeyError: 'frames_tracked_per_obj' #219", "user": { "login": "gamechanger-creator", "id": 222626544, "node_id": "U_kgDODUUC8A", "avatar_url": "https://avatars.githubusercontent.com/u/222626544?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gamechanger-creator", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-10-17T08:22:33Z", "updated_at": "2025-10-26T12:17:17Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "python ./wan/modules/animate/preprocess/preprocess_data.py --ckpt_path ./Wan2.2-Animate-14B/process_checkpoint --video_path ./examples/wan_animate/replace/video.mp4 --refer_path ./examples/wan_animate/replace/image.jpeg --save_path ./examples/wan_animate/replace/process_results --resolution_area 1280 720 --iterations 3 --k 7 --w_len 1 --h_len 1 --replace_flag\n{'ckpt_path': './Wan2.2-Animate-14B/process_checkpoint', 'video_path': './examples/wan_animate/replace/video.mp4', 'refer_path': './examples/wan_animate/replace/image.jpeg', 'save_path': './examples/wan_animate/replace/process_results', 'resolution_area': [1280, 720], 'fps': 30, 'replace_flag': True, 'retarget_flag': False, 'use_flux': False, 'iterations': 3, 'k': 7, 'w_len': 1, 'h_len': 1}\nframe_num: 106\nvideo_fps: 30.0\nfps: 30\ntarget_num: 106\n2025-10-17 08:20:44.839 | INFO | process_pipepline:__call__:67 - Processing pose meta\n2025-10-17 08:20:51.165 | INFO | process_pipepline:__call__:82 - Processing reference image: ./examples/wan_animate/replace/image.jpeg\n2025-10-17 08:20:51.209 | INFO | process_pipepline:__call__:89 - Processing template video: ./examples/wan_animate/replace/video.mp4\nvideo frame: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 106/106 [00:02<00:00, 39.27it/s]\nTraceback (most recent call last):\n File \"/home/ubuntu/Wan2.2/./wan/modules/animate/preprocess/preprocess_data.py\", line 109, in \n process_pipeline(video_path=args.video_path, \n File \"/home/ubuntu/Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 97, in __call__\n masks = self.get_mask(frames, 400, tpl_pose_metas)\n File \"/home/ubuntu/Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 319, in get_mask\n self.predictor.reset_state(inference_state)\n File \"/home/ubuntu/miniconda3/envs/wan/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 116, in decorate_context\n return func(*args, **kwargs)\n File \"/home/ubuntu/Wan2.2/sam2/sam2/sam2_video_predictor.py\", line 678, in reset_state\n self._reset_tracking_results(inference_state)\n File \"/home/ubuntu/Wan2.2/sam2/sam2/sam2_video_predictor.py\", line 701, in _reset_tracking_results\n for v in inference_state[\"frames_tracked_per_obj\"].values():\nKeyError: 'frames_tracked_per_obj'", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/239/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/238", "id": 3524585293, "node_id": "I_kwDOPT-eoM7SFOdN", "number": 238, "title": "Does this model support distributed multi-machine operation?", "user": { "login": "xiaolin8", "id": 9527974, "node_id": "MDQ6VXNlcjk1Mjc5NzQ=", "avatar_url": "https://avatars.githubusercontent.com/u/9527974?v=4", "gravatar_id": "", "url": "https://api.github.com/users/xiaolin8", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-17T05:33:02Z", "updated_at": "2025-10-17T05:33:02Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "torchrun is only for single-machine use; if it cannot run in a distributed manner and utilize GPUs across multiple machines, video generation will still be very, very slow.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/238/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/237", "id": 3521478968, "node_id": "I_kwDOPT-eoM7R5YE4", "number": 237, "title": "Inquiry About Sudden Drop in Animation Quality for WAN2.2 Model in freepik", "user": { "login": "ibraheemaivideos", "id": 238470862, "node_id": "U_kgDODjbGzg", "avatar_url": "https://avatars.githubusercontent.com/u/238470862?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ibraheemaivideos", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-16T11:32:04Z", "updated_at": "2025-10-16T11:32:04Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I am writing to ask about an issue I recently experienced with the WAN2.2 animation model in freepik platform. Until very recently, this model was generating excellent quality animations for my images. However, there has been a sudden and significant drop in the quality of animated results—images now appear much worse than before.\n\nCould you please clarify if there has been an update or change made to the WAN2.2 model? Is there a current issue affecting output quality, or should I expect this to be fixed soon?\n\nI want to clarify that I am not encountering any error messages—the WAN2.2 model is running and producing results every time. However, the problem is that, starting suddenly two days ago, the animation quality became very poor. This drop in quality is consistent and happens with every image I try.\n\nI have tested various animation prompts, multiple different images, and used different browsers (including incognito/private mode), but the issue remains unchanged across all attempts.\n\nThis is surprising because since I joined Freepik, the WAN2.2 model has always produced excellent results and worked perfectly. The sudden change two days ago is very unusual, and I am unable to identify any reason for it.\n\n\nThank you for your support. I look forward to your response, as this feature is vital for my workflow.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/237/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/236", "id": 3521180353, "node_id": "I_kwDOPT-eoM7R4PLB", "number": 236, "title": "what's the base model of wan-animate?", "user": { "login": "renrenzsbbb", "id": 30514838, "node_id": "MDQ6VXNlcjMwNTE0ODM4", "avatar_url": "https://avatars.githubusercontent.com/u/30514838?v=4", "gravatar_id": "", "url": "https://api.github.com/users/renrenzsbbb", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-16T10:03:26Z", "updated_at": "2025-11-06T07:43:06Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Thanks for your great work. I want to know what is the base model of wan-animate? wan2.1 or wan2.2. I can't find the seperately low and high noise model in wan-animate?", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/236/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/235", "id": 3520315696, "node_id": "I_kwDOPT-eoM7R08Ew", "number": 235, "title": "Erzeuge ein Video in dem ein Gulfblauer Porsche 911 Turbo S auf dem Ascari Circuit durch die Schikane fährt.", "user": { "login": "ohilger-del", "id": 238393933, "node_id": "U_kgDODjWaTQ", "avatar_url": "https://avatars.githubusercontent.com/u/238393933?v=4", "gravatar_id": "", "url": "https://api.github.com/users/ohilger-del", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-16T04:54:40Z", "updated_at": "2025-10-16T04:54:40Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/235/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/234", "id": 3520144426, "node_id": "I_kwDOPT-eoM7R0SQq", "number": 234, "title": "Erzeuge ein kurzvideo in dem ein Taucher auf einer Klippe sitzt und in den Sonnenuntergang schaut", "user": { "login": "harrolackmann-stack", "id": 238381698, "node_id": "U_kgDODjVqgg", "avatar_url": "https://avatars.githubusercontent.com/u/238381698?v=4", "gravatar_id": "", "url": "https://api.github.com/users/harrolackmann-stack", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-16T03:04:49Z", "updated_at": "2025-10-16T03:04:49Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Erzeuge ein kurzvideo in dem ein Taucher auf einer Klippe sitzt und in den Sonnenuntergang schaut. ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/234/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/233", "id": 3516128773, "node_id": "I_kwDOPT-eoM7Rk94F", "number": 233, "title": "wan2.2 i2v 14B，图片生成视频，是以输入图片作为首帧吗？实际生成的视频发现首帧与输入图片不完全一样", "user": { "login": "fansticOne", "id": 158262464, "node_id": "U_kgDOCW7kwA", "avatar_url": "https://avatars.githubusercontent.com/u/158262464?v=4", "gravatar_id": "", "url": "https://api.github.com/users/fansticOne", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-15T02:43:20Z", "updated_at": "2025-10-20T06:56:45Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "wan2.2 i2v 14B，图片生成视频，是以输入图片作为首帧吗？实际生成的视频发现首帧与输入图片不完全一样，求解，谢谢", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/233/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 1 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/232", "id": 3513545961, "node_id": "I_kwDOPT-eoM7RbHTp", "number": 232, "title": "wan animate生成效果（pose控制）出现画面模糊和背景异常变化问题", "user": { "login": "Owen1234560", "id": 35532273, "node_id": "MDQ6VXNlcjM1NTMyMjcz", "avatar_url": "https://avatars.githubusercontent.com/u/35532273?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Owen1234560", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-10-14T11:44:44Z", "updated_at": "2025-12-03T15:28:09Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "大家有没有遇到，wan animate生成效果（pose控制）出现画面模糊和背景异常变化问题？还是哪里参数没有设置好。", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/232/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/231", "id": 3513526267, "node_id": "I_kwDOPT-eoM7RbCf7", "number": 231, "title": "Wan2.2-I2V-A14B 图生视频的帧宽高与输入图片的宽高ratio不一致", "user": { "login": "fansticOne", "id": 158262464, "node_id": "U_kgDOCW7kwA", "avatar_url": "https://avatars.githubusercontent.com/u/158262464?v=4", "gravatar_id": "", "url": "https://api.github.com/users/fansticOne", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-14T11:38:01Z", "updated_at": "2025-10-15T01:38:25Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "请教一下，图生视频的帧宽高与输入图片的宽高ratio不一致，希望视频的帧宽高与输入图片的宽高一致，这个size的设置有什么特别注意的吗？感谢！", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/231/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/230", "id": 3512539126, "node_id": "I_kwDOPT-eoM7RXRf2", "number": 230, "title": "[Bug] Potential bug in VAE", "user": { "login": "zyksir", "id": 30949016, "node_id": "MDQ6VXNlcjMwOTQ5MDE2", "avatar_url": "https://avatars.githubusercontent.com/u/30949016?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zyksir", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-14T06:26:26Z", "updated_at": "2025-10-14T08:22:50Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "This is the code in `vae_2_1.py` and this code looks confusing. If `feat_cache[idx] != 'Rep'`, the code is actually doing repeative padding while if `feat_cache[idx] == 'Rep'`, it is doing zero padding. Is this a bug?\n\n```python\nif cache_x.shape[2] < 2 and feat_cache[\n idx] is not None and feat_cache[idx] != 'Rep':\n # cache last frame of last two chunk\n cache_x = torch.cat([\n feat_cache[idx][:, :, -1, :, :].unsqueeze(2).to(\n cache_x.device), cache_x\n ],\n dim=2)\nif cache_x.shape[2] < 2 and feat_cache[\n idx] is not None and feat_cache[idx] == 'Rep':\n cache_x = torch.cat([\n torch.zeros_like(cache_x).to(cache_x.device),\n cache_x\n ],\n dim=2)\n```", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/230/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/229", "id": 3509714695, "node_id": "I_kwDOPT-eoM7RMf8H", "number": 229, "title": "Wan 2.2", "user": { "login": "Horrorfilms1", "id": 153453622, "node_id": "U_kgDOCSWENg", "avatar_url": "https://avatars.githubusercontent.com/u/153453622?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Horrorfilms1", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-13T11:42:44Z", "updated_at": "2025-10-13T11:42:44Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/229/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/228", "id": 3509396909, "node_id": "I_kwDOPT-eoM7RLSWt", "number": 228, "title": "Wan2.2-T2V-A14B模型无法接入Wan2.2-Lightning加速模型", "user": { "login": "HAVELIGHT", "id": 38819199, "node_id": "MDQ6VXNlcjM4ODE5MTk5", "avatar_url": "https://avatars.githubusercontent.com/u/38819199?v=4", "gravatar_id": "", "url": "https://api.github.com/users/HAVELIGHT", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-13T10:20:21Z", "updated_at": "2025-10-13T10:20:21Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Wan2.2-Lightning的readme中提到了调用方法如下，有--lora_dir参数：\ntorchrun --nproc_per_node=8 generate.py --task t2v-A14B --size \"1280*720\" --ckpt_dir ./Wan2.2-T2V-A14B --lora_dir ./Wan2.2-Lightning/Wan2.2-T2V-A14B-4steps-lora-rank64-Seko-V1 --dit_fsdp --t5_fsdp --ulysses_size 8 --base_seed 42 --prompt_file examples/prompt_list.txt\n\n但是本仓库的\n\n[generate.py](https://github.com/user-attachments/files/22882258/generate.py)脚本没有接收该参数。\n请问是还没有开发完成吗？", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/228/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/227", "id": 3508421628, "node_id": "I_kwDOPT-eoM7RHkP8", "number": 227, "title": "关于import flash_attn_interface", "user": { "login": "CHIzhP", "id": 87708148, "node_id": "MDQ6VXNlcjg3NzA4MTQ4", "avatar_url": "https://avatars.githubusercontent.com/u/87708148?v=4", "gravatar_id": "", "url": "https://api.github.com/users/CHIzhP", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-13T04:32:24Z", "updated_at": "2026-01-08T08:06:35Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "### wan/modules/attention.py\n\n```\ntry:\n import flash_attn_interface\n FLASH_ATTN_3_AVAILABLE = True\nexcept ModuleNotFoundError:\n FLASH_ATTN_3_AVAILABLE = False\n```\n\n请问这个`flash_attn_interface`的安装方法以及版本，谢谢！", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/227/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/225", "id": 3506754840, "node_id": "I_kwDOPT-eoM7RBNUY", "number": 225, "title": "wan 2.2的“ 文‘生视频，能否加一个语音声音模块，直接驱动视频产生同步的口型", "user": { "login": "woshilikewang", "id": 199223597, "node_id": "U_kgDOC9_pLQ", "avatar_url": "https://avatars.githubusercontent.com/u/199223597?v=4", "gravatar_id": "", "url": "https://api.github.com/users/woshilikewang", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-12T07:02:15Z", "updated_at": "2025-10-12T07:02:15Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "在wan2.2的文生视频的过程中，导入音频格式的语音“直接驱动”画面人物同步口型，而不是先生成画面再进行口型匹配。\n\n具体来说，是希望找到一种方法或技术，能够在文生视频工作流中，将语音信息与视频生成过程深度融合，使得生成的视频人物能够根据输入的语音实时、自然地做出相应匹配的说话“口型”。", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/225/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/224", "id": 3506651513, "node_id": "I_kwDOPT-eoM7RA0F5", "number": 224, "title": "Saurav singh", "user": { "login": "daparkha88-gif", "id": 237525056, "node_id": "U_kgDODihYQA", "avatar_url": "https://avatars.githubusercontent.com/u/237525056?v=4", "gravatar_id": "", "url": "https://api.github.com/users/daparkha88-gif", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-12T04:14:09Z", "updated_at": "2025-10-12T04:14:09Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "- Video", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/224/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/223", "id": 3506650859, "node_id": "I_kwDOPT-eoM7RAz7r", "number": 223, "title": "Saurav 123", "user": { "login": "daparkha88-gif", "id": 237525056, "node_id": "U_kgDODihYQA", "avatar_url": "https://avatars.githubusercontent.com/u/237525056?v=4", "gravatar_id": "", "url": "https://api.github.com/users/daparkha88-gif", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-12T04:13:09Z", "updated_at": "2025-10-12T04:13:09Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "1. ### ### ", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/223/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/221", "id": 3502063746, "node_id": "I_kwDOPT-eoM7QvUCC", "number": 221, "title": "Wan 2.2 Animate: Question about loss computation on noisy reference and temporal latents.", "user": { "login": "0xLDF", "id": 64844447, "node_id": "MDQ6VXNlcjY0ODQ0NDQ3", "avatar_url": "https://avatars.githubusercontent.com/u/64844447?v=4", "gravatar_id": "", "url": "https://api.github.com/users/0xLDF", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-10-10T09:24:53Z", "updated_at": "2025-10-10T09:24:53Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi Wan team,\nI’m trying to fine-tune Wan-2.2-Animate and would like to clarify how the loss is calculated for the two “extra” latents:\n\nnoisy reference latents\nnoisy temporal latents\nFrom the inference code I see that the denoising network receives three kinds of noisy latents:\n\nnoisy video latents\nnoisy reference latents\nnoisy temporal latents\nThe reference latents is also denoised step-by-step, so their value changes during sampling.\nThis implies that you must have computed a loss on both the reference and the temporal latent during training. Could you share the details?\nDid you simply apply the same flow-matching loss to all three tensors and sum/average them?\n\nFinally, I would like to ask: in current image/video generation models such as Flux-Kontext, Stand-in, Phantom, and RealisDance-DiT, the noisy latents never contain the reference image, and reference is not supervised by any loss during training. What was the motivation behind Wan-2.2-Animate’s decision to add noise to the reference latents.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/221/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/220", "id": 3501691088, "node_id": "I_kwDOPT-eoM7Qt5DQ", "number": 220, "title": "wan2.2-animate-14B 效果特别差", "user": { "login": "huzhongyyuan", "id": 153470133, "node_id": "U_kgDOCSXEtQ", "avatar_url": "https://avatars.githubusercontent.com/u/153470133?v=4", "gravatar_id": "", "url": "https://api.github.com/users/huzhongyyuan", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 4, "created_at": "2025-10-10T07:22:52Z", "updated_at": "2025-12-15T02:00:02Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "跑出来的视频和官方结果差太多了\n模糊性和伪影特别多，一致性也不好\n有一样的吗\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/220/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/219", "id": 3498426096, "node_id": "I_kwDOPT-eoM7Qhb7w", "number": 219, "title": "KeyError: 'frames_tracked_per_obj'", "user": { "login": "XuJianzhi", "id": 30584453, "node_id": "MDQ6VXNlcjMwNTg0NDUz", "avatar_url": "https://avatars.githubusercontent.com/u/30584453?v=4", "gravatar_id": "", "url": "https://api.github.com/users/XuJianzhi", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 5, "created_at": "2025-10-09T09:47:41Z", "updated_at": "2025-10-13T02:49:38Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "(venv_2) root@llmdevelop-46-74c75987bd-zlvmv:/data/try/Wan2.2# python ./wan/modules/animate/preprocess/preprocess_data.py \\\n> --ckpt_path ./Wan2.2-Animate-14B/process_checkpoint \\\n> --video_path ./examples/wan_animate/replace/video.mp4 \\\n> --refer_path ./examples/wan_animate/replace/image.jpeg \\\n> --save_path ./examples/wan_animate/replace/process_results \\\n> --resolution_area 1280 720 \\\n> --iterations 3 \\\n> --k 7 \\\n> --w_len 1 \\\n> --h_len 1 \\\n> --replace_flag\n\n\n\n{'ckpt_path': './Wan2.2-Animate-14B/process_checkpoint', 'video_path': './examples/wan_animate/replace/video.mp4', 'refer_path': './examples/wan_animate/replace/image.jpeg', 'save_path': './examples/wan_animate/replace/process_results', 'resolution_area': [1280, 720], 'fps': 30, 'replace_flag': True, 'retarget_flag': False, 'use_flux': False, 'iterations': 3, 'k': 7, 'w_len': 1, 'h_len': 1}\n/data/try/Wan2.2/venv_2/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py:123: UserWarning: Specified provider 'CUDAExecutionProvider' is not in available provider names.Available providers: 'AzureExecutionProvider, CPUExecutionProvider'\n warnings.warn(\nframe_num: 164\nvideo_fps: 24.0\nfps: 30\ntarget_num: 205\n2025-10-09 05:42:17.053 | INFO | process_pipepline:__call__:67 - Processing pose meta\n2025-10-09 05:43:28.828 | INFO | process_pipepline:__call__:82 - Processing reference image: ./examples/wan_animate/replace/image.jpeg\n2025-10-09 05:43:28.901 | INFO | process_pipepline:__call__:89 - Processing template video: ./examples/wan_animate/replace/video.mp4\nvideo frame: 100%|██████████████████████████| 205/205 [00:07<00:00, 27.23it/s]\nTraceback (most recent call last):\n File \"/data/try/Wan2.2/./wan/modules/animate/preprocess/preprocess_data.py\", line 109, in \n process_pipeline(video_path=args.video_path, \n File \"/data/try/Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 97, in __call__\n masks = self.get_mask(frames, 400, tpl_pose_metas)\n File \"/data/try/Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 319, in get_mask\n self.predictor.reset_state(inference_state)\n File \"/data/try/Wan2.2/venv_2/lib/python3.10/site-packages/torch/utils/_contextlib.py\", line 120, in decorate_context\n return func(*args, **kwargs)\n File \"/data/try/Wan2.2/venv_2/lib/python3.10/site-packages/sam2/sam2_video_predictor.py\", line 678, in reset_state\n self._reset_tracking_results(inference_state)\n File \"/data/try/Wan2.2/venv_2/lib/python3.10/site-packages/sam2/sam2_video_predictor.py\", line 701, in _reset_tracking_results\n for v in inference_state[\"frames_tracked_per_obj\"].values():\nKeyError: 'frames_tracked_per_obj'\n\n\n-------------------\n(venv_2) root@llmdevelop-46-74c75987bd-zlvmv:/data/try/Wan2.2# pip list\nPackage Version\n------------------------ -----------\naccelerate 1.10.1\naiohappyeyeballs 2.6.1\naiohttp 3.13.0\naiosignal 1.4.0\nantlr4-python3-runtime 4.9.3\nasync-timeout 5.0.1\nattrs 25.4.0\naudioread 3.0.1\nbeautifulsoup4 4.14.2\ncertifi 2025.10.5\ncffi 2.0.0\ncharset-normalizer 3.4.3\ncoloredlogs 15.0.1\nconformer 0.3.2\ncontourpy 1.3.2\ncryptography 46.0.2\ncycler 0.12.1\ndashscope 1.24.6\ndecorator 5.2.1\ndecord 0.6.0\ndiffusers 0.35.1\neasydict 1.13\neinops 0.8.1\nfilelock 3.20.0\nflatbuffers 25.9.23\nfonttools 4.60.1\nfrozenlist 1.8.0\nfsspec 2025.9.0\nftfy 6.3.1\ngdown 5.2.0\ngitdb 4.0.12\nGitPython 3.1.45\nhf-xet 1.1.10\nhuggingface-hub 0.35.3\nhumanfriendly 10.0\nhydra-core 1.3.2\nHyperPyYAML 1.2.2\nidna 3.10\nimageio 2.37.0\nimageio-ffmpeg 0.6.0\nimportlib_metadata 8.7.0\ninflect 7.5.0\niopath 0.1.10\nJinja2 3.1.6\njoblib 1.5.2\nkaldifst 1.7.17\nkiwisolver 1.4.9\nlazy_loader 0.4\nlibrosa 0.11.0\nlightning 2.5.5\nlightning-utilities 0.15.2\nllvmlite 0.45.1\nloguru 0.7.3\nmarkdown-it-py 4.0.0\nMarkupSafe 3.0.3\nmatplotlib 3.10.7\nmdurl 0.1.2\nmodelscope 1.30.0\nmore-itertools 10.8.0\nmoviepy 2.2.1\nmpmath 1.3.0\nmsgpack 1.1.2\nmultidict 6.7.0\nnetworkx 3.4.2\nnumba 0.62.1\nnumpy 1.26.4\nnvidia-cublas-cu12 12.8.4.1\nnvidia-cuda-cupti-cu12 12.8.90\nnvidia-cuda-nvrtc-cu12 12.8.93\nnvidia-cuda-runtime-cu12 12.8.90\nnvidia-cudnn-cu12 9.10.2.21\nnvidia-cufft-cu12 11.3.3.83\nnvidia-cufile-cu12 1.13.1.3\nnvidia-curand-cu12 10.3.9.90\nnvidia-cusolver-cu12 11.7.3.90\nnvidia-cusparse-cu12 12.5.8.93\nnvidia-cusparselt-cu12 0.7.1\nnvidia-nccl-cu12 2.27.3\nnvidia-nvjitlink-cu12 12.8.93\nnvidia-nvtx-cu12 12.8.90\nomegaconf 2.3.0\nonnxruntime 1.23.1\nopenai-whisper 20250625\nopencv-python 4.11.0.86\npackaging 25.0\npillow 11.3.0\npip 23.0.1\nplatformdirs 4.5.0\npooch 1.8.2\nportalocker 3.2.0\nproglog 0.1.12\npropcache 0.4.1\nprotobuf 6.32.1\npsutil 7.1.0\npyarrow 21.0.0\npycparser 2.23\nPygments 2.19.2\npyparsing 3.2.5\nPySocks 1.7.1\npython-dateutil 2.9.0.post0\npython-dotenv 1.1.1\npytorch-lightning 2.5.5\npyworld 0.3.5\nPyYAML 6.0.3\nregex 2025.9.18\nrequests 2.32.5\nrich 14.1.0\nruamel.yaml 0.18.15\nruamel.yaml.clib 0.2.14\nsafetensors 0.6.2\nsam2 1.1.0\nscikit-learn 1.7.2\nscipy 1.15.3\nsentencepiece 0.2.1\nsetuptools 65.5.0\nsix 1.17.0\nsmmap 5.0.2\nsoundfile 0.13.1\nsoupsieve 2.8\nsoxr 1.0.0\nsympy 1.14.0\nthreadpoolctl 3.6.0\ntiktoken 0.12.0\ntokenizers 0.21.4\ntorch 2.8.0\ntorchmetrics 1.8.2\ntorchvision 0.23.0\ntqdm 4.67.1\ntransformers 4.51.3\ntriton 3.4.0\ntypeguard 4.4.4\ntyping_extensions 4.15.0\nurllib3 2.5.0\nwcwidth 0.2.14\nwebsocket-client 1.9.0\nwetext 0.1.0\nwget 3.2\nyarl 1.22.0\nzipp 3.23.0\n\n", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/219/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/216", "id": 3490080002, "node_id": "I_kwDOPT-eoM7QBmUC", "number": 216, "title": "S2V模型生成的视频，下方总会出现马赛克一样的横幅", "user": { "login": "a63976659", "id": 193756234, "node_id": "U_kgDOC4x8Sg", "avatar_url": "https://avatars.githubusercontent.com/u/193756234?v=4", "gravatar_id": "", "url": "https://api.github.com/users/a63976659", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-07T06:07:06Z", "updated_at": "2025-10-09T03:21:23Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "S2V模型生成的视频，下方总会出现马赛克一样的横幅\nThe videos generated by the S2V model always have a mosaic-like banner appearing at the bottom.\n\n\"Image\"\n\n我使用的是ComfyUI官方模版中的s2v工作流，\nI am using the s2v workflow from the official ComfyUI template.\nS2V 14B bf16模型+4步lora\nS2V 14B bf16 model + 4-step LORA\n\n\"Image\"\n\nhttps://github.com/user-attachments/assets/75fcb5de-c01b-4e81-9228-50b0579776fe", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/216/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/214", "id": 3481679201, "node_id": "I_kwDOPT-eoM7PhjVh", "number": 214, "title": "Community Work: Ovi — Human-centric Video+Audio Generation (video branch initialized from Wan2.2)", "user": { "login": "aaxwaz", "id": 9063786, "node_id": "MDQ6VXNlcjkwNjM3ODY=", "avatar_url": "https://avatars.githubusercontent.com/u/9063786?v=4", "gravatar_id": "", "url": "https://api.github.com/users/aaxwaz", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 4, "created_at": "2025-10-03T16:16:04Z", "updated_at": "2025-10-04T09:07:40Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi Wan team! Thanks a ton for releasing Wan2.2 🙏 it’s been super inspiring for us.\n\nWe just open-sourced Ovi, an audio-video generation model built on top of Wan2.2’s TI2V backbone. On top of that, we trained our 5B audio branch + 1B fusion branch from scratch using our high-quality internal video/audio datasets. Ovi can do synced speech, multi-person dialogue, music, and sound effects.\n\nLinks here:\n\nGitHub: https://github.com/character-ai/Ovi\n\nPaper: https://arxiv.org/abs/2510.01284\n\nWebsite demo: https://character-ai.github.io/Ovi\n\nWould be awesome if you could add Ovi under the Community Works section 🙌", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/214/reactions", "total_count": 1, "+1": 0, "-1": 0, "laugh": 0, "hooray": 1, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/213", "id": 3474562603, "node_id": "I_kwDOPT-eoM7PGZ4r", "number": 213, "title": "thanks", "user": { "login": "noname0121", "id": 22991442, "node_id": "MDQ6VXNlcjIyOTkxNDQy", "avatar_url": "https://avatars.githubusercontent.com/u/22991442?v=4", "gravatar_id": "", "url": "https://api.github.com/users/noname0121", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-10-01T18:04:57Z", "updated_at": "2025-10-07T03:44:22Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "thanks", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/213/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/212", "id": 3469401521, "node_id": "I_kwDOPT-eoM7Oyt2x", "number": 212, "title": "malaikat mikail dengan tugasnya", "user": { "login": "choirulihsani09-create", "id": 235315808, "node_id": "U_kgDODgaiYA", "avatar_url": "https://avatars.githubusercontent.com/u/235315808?v=4", "gravatar_id": "", "url": "https://api.github.com/users/choirulihsani09-create", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-30T13:14:19Z", "updated_at": "2025-09-30T13:14:19Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Malaikat Mikail adalah salah satu malaikat utama yang diutus Allah. Tugas utamanya adalah mengatur rezeki bagi seluruh makhluk, seperti menurunkan hujan, menyuburkan tumbuhan, dan menjaga keseimbangan alam. Mikail digambarkan sebagai malaikat yang penuh kasih sayang terhadap manusia dan selalu mendoakan kebaikan bagi orang-orang beriman. Dalam Islam, ia bekerja bersama malaikat lain sesuai perintah Allah tanpa pernah lelah atau membangkang. Karena jasanya, umat manusia dapat merasakan kenikmatan hidup sehari-hari, mulai dari makanan yang tumbuh di bumi hingga hujan yang menyegarkan.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/212/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/211", "id": 3467411475, "node_id": "I_kwDOPT-eoM7OrIAT", "number": 211, "title": "Poor generation quality when using diffusers", "user": { "login": "INV-WZQ", "id": 33061675, "node_id": "MDQ6VXNlcjMzMDYxNjc1", "avatar_url": "https://avatars.githubusercontent.com/u/33061675?v=4", "gravatar_id": "", "url": "https://api.github.com/users/INV-WZQ", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 2, "created_at": "2025-09-30T03:48:30Z", "updated_at": "2025-10-23T07:30:34Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I encountered a problem with video generation using the Wan2.2-TI2V-5B-Diffusers model with diffusers. The results show poor quality. Below are the code, environment details, and the results. \n\n## Code\n```\nimport sys\nimport time\nimport torch\nfrom diffusers.utils import export_to_video\nfrom diffusers import AutoencoderKLWan, WanPipeline\nfrom diffusers.schedulers.scheduling_unipc_multistep import UniPCMultistepScheduler\n\nmodel_id = \"Wan-AI/Wan2.2-TI2V-5B-Diffusers\"\n\nvae = AutoencoderKLWan.from_pretrained(model_id, subfolder=\"vae\", torch_dtype=torch.bfloat16)\n\nflow_shift = 5.0 \npipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)\npipe.scheduler.config.flow_shift = flow_shift\nscheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)\npipe.scheduler = scheduler\npipe.to(\"cuda\")\n\nprompt = \"A cat and a dog baking a cake together in a kitchen. The cat is carefully measuring flour, while the dog is stirring the batter with a wooden spoon. The kitchen is cozy, with sunlight streaming through the window.\"\nnegative_prompt = \"Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards\"\n\noutput = pipe(\n prompt=prompt,\n negative_prompt=negative_prompt,\n height=704,\n width=1280,\n num_frames=17,\n guidance_scale=5.0,\n ).frames[0]\nexport_to_video(output, \"output.mp4\", fps=16)\n```\n\n## Environment\n```\ntorch==2.4.1+cu121\ndiffusers==0.35.1\ntransformers==4.49.0\n```\n\n## Result\n\nhttps://github.com/user-attachments/assets/73002f52-ef54-46d8-b45a-32e4823f3377", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/211/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/209", "id": 3466133878, "node_id": "I_kwDOPT-eoM7OmQF2", "number": 209, "title": "ModuleNotFoundError: No module named 'moviepy'", "user": { "login": "msdsm", "id": 66203251, "node_id": "MDQ6VXNlcjY2MjAzMjUx", "avatar_url": "https://avatars.githubusercontent.com/u/66203251?v=4", "gravatar_id": "", "url": "https://api.github.com/users/msdsm", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-29T18:31:06Z", "updated_at": "2025-09-29T18:31:06Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "## Description\n\nThe `requirements_animate.txt` file is missing the `moviepy` dependency, which causes a `ModuleNotFoundError` when running the animation preprocessing script.\n\n## Error Message\nModuleNotFoundError: No module named 'moviepy'\n\n## Steps to Reproduce\n\n1. Install dependencies from `requirements_animate.txt`:\n```bash\npip install -r requirements_animate.txt\n```\n\n2. Run the preprocessing script:\n```bash\npython ./wan/modules/animate/preprocess/preprocess_data.py \\\n --ckpt_path ./Wan2.2-Animate-14B/process_checkpoint \\\n --video_path ./examples/wan_animate/animate/video.mp4 \\\n --refer_path ./examples/wan_animate/animate/image.jpeg \\\n --save_path ./examples/wan_animate/animate/process_results \\\n --resolution_area 1280 720 \\\n --retarget_flag \\\n --use_flux\n```", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/209/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/208", "id": 3465404759, "node_id": "I_kwDOPT-eoM7OjeFX", "number": 208, "title": "Wan 2.2 Animate API???", "user": { "login": "afiq90", "id": 8640517, "node_id": "MDQ6VXNlcjg2NDA1MTc=", "avatar_url": "https://avatars.githubusercontent.com/u/8640517?v=4", "gravatar_id": "", "url": "https://api.github.com/users/afiq90", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-09-29T14:56:57Z", "updated_at": "2025-12-03T15:29:08Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi, out of curiosity does Wan official provide API for Wan 2.2 Animate??", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/208/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/207", "id": 3464013891, "node_id": "I_kwDOPT-eoM7OeKhD", "number": 207, "title": "wan2.2 s2v生成的跳舞视频这么抽象，是什么原因？", "user": { "login": "gangxu822", "id": 24646369, "node_id": "MDQ6VXNlcjI0NjQ2MzY5", "avatar_url": "https://avatars.githubusercontent.com/u/24646369?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gangxu822", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-09-29T09:17:44Z", "updated_at": "2025-10-10T07:24:26Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "生成视频：\n\nhttps://github.com/user-attachments/assets/6fbb1089-f941-41c3-9dc5-0f82f3720d2a\n生成pose视频的原视频：\n\nhttps://github.com/user-attachments/assets/df2d04c2-866f-4e5c-ab37-188ce418ce88\npose视频：\n\nhttps://github.com/user-attachments/assets/8f265a91-d2ec-4f3f-a7c2-08ec1fe195e0", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/207/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/206", "id": 3463875475, "node_id": "I_kwDOPT-eoM7OdouT", "number": 206, "title": "Wan-Animate 长视频，在20多秒就会，严重变形", "user": { "login": "SevenMpp", "id": 34155680, "node_id": "MDQ6VXNlcjM0MTU1Njgw", "avatar_url": "https://avatars.githubusercontent.com/u/34155680?v=4", "gravatar_id": "", "url": "https://api.github.com/users/SevenMpp", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 4, "created_at": "2025-09-29T08:43:51Z", "updated_at": "2025-12-12T09:15:41Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Wan-Animate 长视频，共计5分钟的视频，跑了2天的时间，但是在25秒左右开始就会，严重变形，色彩和颜色都出现了很大的情况\n\n\n由于文件太大，无法上传，目前 视频的对比度 严重下降，并且视频种的任务，严重变形", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/206/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/203", "id": 3462952994, "node_id": "I_kwDOPT-eoM7OaHgi", "number": 203, "title": "wan2.2 animate的输入图像比例都要求16:9嘛？", "user": { "login": "gangxu822", "id": 24646369, "node_id": "MDQ6VXNlcjI0NjQ2MzY5", "avatar_url": "https://avatars.githubusercontent.com/u/24646369?v=4", "gravatar_id": "", "url": "https://api.github.com/users/gangxu822", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-29T02:14:04Z", "updated_at": "2025-09-29T02:14:04Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "命令中的resolution_area是不是也得是16：9比例？\n\n\"Image\"", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/203/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/201", "id": 3461983008, "node_id": "I_kwDOPT-eoM7OWasg", "number": 201, "title": "replacement任务处理超过超过30s的视频出现漂移问题", "user": { "login": "Owenmessizz", "id": 919492, "node_id": "MDQ6VXNlcjkxOTQ5Mg==", "avatar_url": "https://avatars.githubusercontent.com/u/919492?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Owenmessizz", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-28T14:31:14Z", "updated_at": "2025-09-28T14:31:14Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "用官方文档处理replacement任务的时候会出现视频越到后面颜色越深画面也越来越模糊的问题，我尝试去掉了use_relighting_lora选项，结果出来的视频颜色倒没有变深了，但是整个视频会很晃，想问下这个是什么原因呢？是素材的问题还是说目前处理长视频的replacement任务还存在缺陷", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/201/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/200", "id": 3461494082, "node_id": "I_kwDOPT-eoM7OUjVC", "number": 200, "title": "Wan 2.2 Animate: Question about loss computation on noisy reference and temporal latents.", "user": { "login": "0xLDF", "id": 64844447, "node_id": "MDQ6VXNlcjY0ODQ0NDQ3", "avatar_url": "https://avatars.githubusercontent.com/u/64844447?v=4", "gravatar_id": "", "url": "https://api.github.com/users/0xLDF", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-09-28T09:52:27Z", "updated_at": "2025-09-28T10:20:59Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi Wan team,\nI’m trying to fine-tune Wan-2.2-Animate and would like to clarify how the loss is calculated for the two “extra” latents:\n- noisy reference latents\n- noisy temporal latents\n\nFrom the inference code I see that the denoising network receives three kinds of noisy latents:\n- noisy video latents\n- noisy reference latents\n- noisy temporal latents\n\nThe reference latents is also denoised step-by-step, so their value changes during sampling.\nThis implies that you must have computed a loss on both the reference and the temporal latent during training. Could you share the details?\nDid you simply apply the same flow-matching loss to all three tensors and sum/average them?\n\nFinally, I would like to ask: in current image/video generation models such as Flux-Kontext, Stand-in, Phantom, and RealisDance-DiT, the noisy latents never contain the reference image, and reference is not supervised by any loss during training. What was the motivation behind Wan-2.2-Animate’s decision to add noise to the reference latents.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/200/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/199", "id": 3461408734, "node_id": "I_kwDOPT-eoM7OUOfe", "number": 199, "title": "SAM2Base has no attribute init_state_v2", "user": { "login": "imneonizer", "id": 13932346, "node_id": "MDQ6VXNlcjEzOTMyMzQ2", "avatar_url": "https://avatars.githubusercontent.com/u/13932346?v=4", "gravatar_id": "", "url": "https://api.github.com/users/imneonizer", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-09-28T09:01:43Z", "updated_at": "2025-10-10T00:23:29Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "I see that `process_pipeline.py` has this function `self.predictor.init_state_v2(frames=each_frames)` but no such function is supported by SAM2 from where did you guys installed SAM2 because in the requirements file i see `git+https://github.com/facebookresearch/sam2.git@0e78a118995e66bb27d78518c4bd9a3e95b4e266#egg=SAM-2 ` which is not helpful.", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/199/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/196", "id": 3461011976, "node_id": "I_kwDOPT-eoM7OStoI", "number": 196, "title": "inference i2v-A14B on 2*A100-80G OOM", "user": { "login": "Shmily-YH", "id": 51698248, "node_id": "MDQ6VXNlcjUxNjk4MjQ4", "avatar_url": "https://avatars.githubusercontent.com/u/51698248?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Shmily-YH", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 5, "created_at": "2025-09-28T03:30:03Z", "updated_at": "2025-12-17T10:43:03Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "这个问题很奇怪哦，单卡运行不会报错，两卡就会报错OOM\n\n单卡命令：\npython generate.py --task i2v-A14B --size 1280*720 --ckpt_dir ./Wan2.2-I2V-A14B-BF16 --offload_model True --convert_model_dtype --image examples/i2v_input.JPG --prompt\n\n双卡命令：\ntorchrun --nproc_per_node=2 generate.py --task i2v-A14B --size 1280*720 --offload_model True --convert_model_dtype --t5_cpu --ckpt_dir ./Wan2.2-I2V-A14B-BF16 --image examples/i2v_input.JPG --dit_fsdp --t5_fsdp --ulysses_size 2 --prompt", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/196/reactions", "total_count": 5, "+1": 3, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 2 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/195", "id": 3460190069, "node_id": "I_kwDOPT-eoM7OPk91", "number": 195, "title": "Wan2.5 循环视频支持", "user": { "login": "zwukong", "id": 39182231, "node_id": "MDQ6VXNlcjM5MTgyMjMx", "avatar_url": "https://avatars.githubusercontent.com/u/39182231?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zwukong", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 5, "created_at": "2025-09-27T11:46:10Z", "updated_at": "2025-12-19T10:46:14Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "一个视频可能需要首帧与尾帧相同，目前单纯通过单图+提示词是很难做到的。尾帧可以做到相似但很难做到完全相同。我看即梦有多帧功能来实现，wan目前还没有看到这个功能。首尾帧与多帧还是不一样的，多帧是一个提示词，多帧拆分为多个镜头", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/195/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/194", "id": 3460185000, "node_id": "I_kwDOPT-eoM7OPjuo", "number": 194, "title": "Wan2.5 多人时嘴型混乱", "user": { "login": "zwukong", "id": 39182231, "node_id": "MDQ6VXNlcjM5MTgyMjMx", "avatar_url": "https://avatars.githubusercontent.com/u/39182231?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zwukong", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 1, "created_at": "2025-09-27T11:39:56Z", "updated_at": "2025-09-27T18:11:43Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "图生视频与文生视频都能遇到。多人时嘴型可能出现混乱，不该说的也张嘴说。", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/194/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/193", "id": 3460183075, "node_id": "I_kwDOPT-eoM7OPjQj", "number": 193, "title": "Wan2.5多图像输入", "user": { "login": "zwukong", "id": 39182231, "node_id": "MDQ6VXNlcjM5MTgyMjMx", "avatar_url": "https://avatars.githubusercontent.com/u/39182231?v=4", "gravatar_id": "", "url": "https://api.github.com/users/zwukong", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-27T11:38:10Z", "updated_at": "2025-09-27T11:38:10Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "目前因为支持10秒的时长，可能需要拆分为多个镜头。每个镜头里面的人物又可能是不一样的，因此多图像输入感觉是刚需，不知道正式版会不会支持。我试着做一个短片也遇到这个问题，图生视频的话单图是不够的。", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/193/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/191", "id": 3458138182, "node_id": "I_kwDOPT-eoM7OHwBG", "number": 191, "title": "bot", "user": { "login": "muhinbot09-arch", "id": 234644978, "node_id": "U_kgDODfxl8g", "avatar_url": "https://avatars.githubusercontent.com/u/234644978?v=4", "gravatar_id": "", "url": "https://api.github.com/users/muhinbot09-arch", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-26T15:39:03Z", "updated_at": "2025-09-26T15:39:03Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": null, "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/191/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/190", "id": 3458068881, "node_id": "I_kwDOPT-eoM7OHfGR", "number": 190, "title": "Wan Animate: face was covered by hair", "user": { "login": "sweihub", "id": 5334841, "node_id": "MDQ6VXNlcjUzMzQ4NDE=", "avatar_url": "https://avatars.githubusercontent.com/u/5334841?v=4", "gravatar_id": "", "url": "https://api.github.com/users/sweihub", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 0, "created_at": "2025-09-26T15:18:46Z", "updated_at": "2025-09-26T15:23:44Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "用中文说啦，发现Wan Animate有两个问题\n\n1. 脸被头发挡住了（尝试了两次，一样问题），请看视频对比\n2. 音画不同步，以下视频没有加声音，实际生成的视频比原来的视频时间要长点，导致音画不同步，（帧率24，滑动窗口77）。越长的视频差异越明显。\n\nhttps://github.com/user-attachments/assets/bc7fbe33-ada7-43fb-bb6f-463fa1fd7d79", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/190/reactions", "total_count": 0, "+1": 0, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/189", "id": 3453463444, "node_id": "I_kwDOPT-eoM7N16uU", "number": 189, "title": "Wan2.2 Animate", "user": { "login": "FidanVural", "id": 56233156, "node_id": "MDQ6VXNlcjU2MjMzMTU2", "avatar_url": "https://avatars.githubusercontent.com/u/56233156?v=4", "gravatar_id": "", "url": "https://api.github.com/users/FidanVural", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 3, "created_at": "2025-09-25T12:48:49Z", "updated_at": "2025-12-27T04:03:47Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "Hi there,\n\nI'm trying to do preprocessing with the following command:\n\n`python ./wan/modules/animate/preprocess/preprocess_data.py --ckpt_path /mnt/dev-s1/Wan2.2-Animate-14B/process_checkpoint --video_path ./examples/wan_animate/animate/video.mp4 --refer_path ./examples/wan_animate/animate/image.jpeg --save_path ./examples/wan_animate/animate/process_results --resolution_area 1280 720 --retarget_flag --use_flux`\n\nHowever, I'm getting the error below:\n\n{'ckpt_path': '/mnt/dev-s1/Wan2.2-Animate-14B/process_checkpoint', 'video_path': './examples/wan_animate/animate/video.mp4', 'refer_path': './examples/wan_animate/animate/image.jpeg', 'save_path': './examples/wan_animate/animate/process_results', 'resolution_area': [1280, 720], 'fps': 30, 'replace_flag': False, 'retarget_flag': True, 'use_flux': True, 'iterations': 3, 'k': 7, 'w_len': 1, 'h_len': 1}\nTraceback (most recent call last):\n File \"/home/wrusr/Wan2.2/./wan/modules/animate/preprocess/preprocess_data.py\", line 107, in \n process_pipeline = ProcessPipeline(det_checkpoint_path=det_checkpoint_path, pose2d_checkpoint_path=pose2d_checkpoint_path, sam_checkpoint_path=sam2_checkpoint_path, flux_kontext_path=flux_kontext_path)\n File \"/home/wrusr/Wan2.2/wan/modules/animate/preprocess/process_pipepline.py\", line 30, in __init__\n self.pose2d = Pose2d(checkpoint=pose2d_checkpoint_path, detector_checkpoint=det_checkpoint_path)\n File \"/home/wrusr/Wan2.2/wan/modules/animate/preprocess/pose2d.py\", line 354, in __init__\n self.model = ViTPose(checkpoint, device)\n File \"/home/wrusr/Wan2.2/wan/modules/animate/preprocess/pose2d.py\", line 311, in __init__\n super(ViTPose, self).__init__(checkpoint, device=device)\n File \"/home/wrusr/Wan2.2/wan/modules/animate/preprocess/pose2d.py\", line 36, in __init__\n self.session = onnxruntime.InferenceSession(checkpoint,\n File \"/home/wrusr/miniconda3/envs/wan2_2_animate_env/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py\", line 472, in __init__\n self._create_inference_session(providers, provider_options, disabled_optimizers)\n File \"/home/wrusr/miniconda3/envs/wan2_2_animate_env/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py\", line 561, in _create_inference_session\n sess.initialize_session(providers, provider_options, disabled_optimizers)\nonnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : Deserialize tensor backbone.blocks.16.norm1.bias failed.tensorprotoutils.cc:1091 GetExtDataFromTensorProto External initializer: backbone.blocks.16.norm1.bias offset: 0 size to read: 5120 given file_length: 129 are out of bounds or can not be read in full.\n\n\nCan you please help me this error?", "closed_by": null, "reactions": { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/189/reactions", "total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0 }, "performed_via_github_app": null, "state_reason": null, "linked_prs": [] }, { "url": "https://api.github.com/repos/Wan-Video/Wan2.2/issues/188", "id": 3452485618, "node_id": "I_kwDOPT-eoM7NyL_y", "number": 188, "title": "[Wan2.2 Animate] 本地部署跑的效果和官方Demo效果差距较大", "user": { "login": "TwinterT", "id": 42286859, "node_id": "MDQ6VXNlcjQyMjg2ODU5", "avatar_url": "https://avatars.githubusercontent.com/u/42286859?v=4", "gravatar_id": "", "url": "https://api.github.com/users/TwinterT", "type": "User", "user_view_type": "public", "site_admin": false }, "labels": {}, "state": "open", "locked": false, "assignee": null, "assignees": {}, "milestone": null, "comments": 5, "created_at": "2025-09-25T08:15:16Z", "updated_at": "2025-12-09T06:59:30Z", "closed_at": null, "author_association": "NONE", "type": null, "active_lock_reason": null, "sub_issues_summary": { "total": 0, "completed": 0, "percent_completed": 0 }, "issue_dependencies_summary": { "blocked_by": 0, "total_blocked_by": 0, "blocking": 0, "total_blocking": 0 }, "body": "RT，本地部署的效果和官方Demo(ModelScope)上的效果差距比较明显。\n想问下\n1. 官方Demo在模型上和开源的模型一样的吗，还是有内部优化版本。\n2. 如果Demo和开源用的模型是一样的，前处理或后处理流程和开源方案有什么不同呢。\n\n以下是我本地部署和官方效果的对比：\n官方效果：\n