ashawkey / RAD-NeRF

Real-time Neural Radiance Talking Portrait Synthesis via Audio-spatial Decomposition
MIT License
878 stars 153 forks source link

questions about head pose #16

Open WuNvRenQiaoDaiMaZiRanShen opened 1 year ago

WuNvRenQiaoDaiMaZiRanShen commented 1 year ago

Different drive audio from the same person share the same headpose.json,is it normal?

ashawkey commented 1 year ago

@WuNvRenQiaoDaiMaZiRanShen Hi, currently we just repeat the head poses in training sequence. It's better to predict a head pose sequence based on audio, but this is not done in this work.