lifeplayer

The performance of detecting shot changes is poor

6

**Problem/Use Case** I found that the performance of detecting shot changes using PySceneDetect is not as effective as Adobe Premiere Pro. Could you please provide a parameter adjustment solution to...

The audio-driven video image is flickering

3

Thank you for your excellent work! The flickering in the demo video driven by audio is quite severe. Is there any solution to optimize it?

The fidelity of the generated faces is poor

2

Thank your great work! I conducted some experiments and found that the fidelity of the generated faces is poor, generated person does not resemble the original video closely. Could you...

https://github.com/taishan1994/qlora-chinese-LLM/blob/b9495bfc3c74188054b9d6c1b8fa26ceb8e81b20/chat.py#L118 同样的peft版本，但是会报错。您这边的环境能跑通么？ 另外，删去这个参数就可以跑通了

关于流式输出的问题

6

感谢作者开源代码，我试用了下musetalk的，好像并没有实现实时对话的效果，流式体现在哪里

In `create_tracked_mask` and `create_tracked_mask` function code, mouth_detector is actually a face detector, however, the detector often fail to detect faces.

 The specific situation is shown in the figure. What are the possible solutions?

The refine_foreground function takes a long time to execute; how can it be optimized?

17

When executing the demo, I found that `refine_foreground` takes more than 4 seconds. Is there any way to speed up this process?