search

Zz-ww / SadTalker-Video-Lip-Sync

本项目基于SadTalkers实现视频唇形合成的Wav2lip。通过以视频文件方式进行语音驱动生成唇形，设置面部区域可配置的增强方式进行合成唇形（人脸）区域画面增强，提高生成唇形的清晰度。使用DAIN 插帧的DL算法对生成视频进行补帧，补充帧间合成唇形的动作过渡，使合成的唇形更为流畅、真实以及自然。
1.88k stars 324 forks source link

Adjust 'audio2exp' to less than every 10 frames for people who talk fast #7

Open bheestandLMT opened 1 year ago

bheestandLMT commented 1 year ago

Thanks for your work on this - sorry this is in English. The results are very intriguing using this repository but on my tests, the lips are not moving correctly. I wonder if it is due to the audio only being checked every 10 frames.

image

Also, would be nice to have the frame rate of the video parsed from the input video and used instead of assuming 25fps.

Looking forward to your response (you can respond in Chinese, I will use a translation app.

Zz-ww commented 1 year ago

Thanks for your work on this - sorry this is in English. The results are very intriguing using this repository but on my tests, the lips are not moving correctly. I wonder if it is due to the audio only being checked every 10 frames. image

Also, would be nice to have the frame rate of the video parsed from the input video and used instead of assuming 25fps.

Looking forward to your response (you can respond in Chinese, I will use a translation app.

There is a bug in the project. I'm working on it.