You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I use nearly 4w high quality talking head data to train wav2lip288x288. I found that during training, the generated bottom half face always blur. I try to use GAN loss and perceptual loss, but it doesn't help. What should I do. Is the wav2lip model structure has limitation on image quality? Or any loss can improve this?
The text was updated successfully, but these errors were encountered:
I use nearly 4w high quality talking head data to train wav2lip288x288. I found that during training, the generated bottom half face always blur. I try to use GAN loss and perceptual loss, but it doesn't help. What should I do. Is the wav2lip model structure has limitation on image quality? Or any loss can improve this?
The text was updated successfully, but these errors were encountered: