No lip variation in the generated frames.

[PrashanthaTP]

Sync and sequence discriminator don't seem to learn

[PrashanthaTP]

For sync disriminator

• it should be trained seperately for x epochs till generator can generate fair images of face
• it helps sync disc to focus on lips then as they are clearly observed in the image.

[PrashanthaTP]

Things that help

• Large dataset with different actors,with equal distribution of genders
• Training sync discriminator seperately initially as done in wav2lip.