Open thatnerdyaigirl opened 3 months ago
Hey @ZiqiaoPeng , this issue is at fixed time frames, no matter if i train with hubert or ave or give any inference audio to test. Tried to train with higher num_rays and patch size but still the same issue
I'm guessing if it happens with any inference audio, the issue is from the preprocessing (e.g. face mask generation) of that particular frame. Maybe check the face mask?
Hi, I tried to finetune it as well with hubert and ave but I keep getting these background dots of the background colour, how can I resolve this?