aigc-apps / aigc-apps/VideoX-Fun
Wan2.2 5B 生成结果异常
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2.3k
- Forks
- 188
- Avg merge
- 2d 22h
- Merged PRs (30d)
- 3
Description
您好！非常感谢您这边对于videoxfun仓库的维护
我最近在基于videoxfun仓库的wan2.2 5B训练和推理代码进行二次开发，
然而在测试过程中我发现videoxfun的wan2.2 5B推理脚本在生成内容时质量相对于官方脚本存在非常大的差距。
以下是我分别使用本仓库的推理代码和Comfyui官方工作流，使用H800基于同一张图进行图生视频(480 x 832 x 121)的结果
下面这个是Comfyui官方工作流的结果
https://github.com/user-attachments/assets/578a3ca1-456e-4a96-a4a7-1eaf8d0aff6b
这个是本仓库的推理代码的结果
https://github.com/user-attachments/assets/3fdd5733-b4b0-445d-bd66-969ee23f1527
可以看到，本仓库的推理代码在生成的视频存在多个片段，并且质量相比Comfyui官方工作流存在较大差距，特别是在3s后会突然换到另一个无关场景，我仔细设置了cfg，shift保持一致并关闭了cpu offload和teacache，问题仍未解决。
我这边附上图片素材，prompt是：相机缓慢向前移动
期待您的回复！
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the repository's Wan2.2 5B image-to-video inference script and reproduce the reported case using the supplied image, prompt, 480 x 832 x 121 resolution, and H800 settings. Compare its configuration and outputs with the ComfyUI official workflow, including CFG, shift, CPU offload, and TeaCache; done means identifying the cause of the scene change and the quality gap.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100