about the TH_1KH dataset

[leegang79]

the dataset downloaded has no audio track. So I downloaded audio stream audioStream = yt.streams.filter(audio_codec='mp4a.40.2', type="audio").first()

then merge the video and audio track via ffmpeg cmd.

I tryed download and processed top 10% videos in train_video_ids.txt. But after video split and crop, there are lots of video clips doesn't contains sync video/audio, for example, the people showing the video is drinking, another one outside of the picture is talking.

Is it possible to share the dataset TH1KH_512 after prepropessed via Baidu Drive? Thanks a lot

[RayShing]

See this issue?

https://github.com/tcwang0509/TalkingHead-1KH/issues/2