What may have caused the extra noise added to a face during inference after adapting the current code framework in liveTalking from ernerf to Synctalk?

ZiqiaoPeng / SyncTalk

[CVPR 2024] This is the official source for our paper "SyncTalk: The Devil is in the Synchronization for Talking Head Synthesis"

https://ziqiaopeng.github.io/synctalk/

Other

1.34k stars 161 forks source link

What may have caused the extra noise added to a face during inference after adapting the current code framework in liveTalking from ernerf to Synctalk? #238

Open EYcab opened 5 days ago

EYcab commented 5 days ago

The inference of my model based on SyncTalk work perfectly fine, but after adapting this model to liveTalking codes based on the ERNERF approach, there was noise added to her face.

Please give me some hints on handling this issue. Thank you in advance.

This is also posted on https://github.com/lipku/LiveTalking/issues/302