search

Weizhi-Zhong / IP_LAP

CVPR2023 talking face implementation for Identity-Preserving Talking Face Generation With Landmark and Appearance Priors
Apache License 2.0
635 stars 70 forks source link

Distinguish face appearance and pose #56

Open alasokolova opened 2 months ago

alasokolova commented 2 months ago

Hey, can i somehow distinguish the appearance and the facial pose of the person? (get landmarks from input video, but the face itself from another source, so that it would work as a face reenactment?)

hatakeyama-poke commented 2 months ago

Rendering network inputs landmark sketch, face image and sound, trains the inpainting of face. So, if you can prepare face images with almost the same face shape and angle as the input sketch, maybe yes.