Pre-trained models of Portrait4D and Portrait4D-v2:
- genhead-ffhq512: GenHead trained on FFHQ dataset at 512x512 resolution
- portrait4d-genhead512: Portrait4D trained with synthetic data of GenHead at 512x512 resolution
- portrait4d-static-genhead512: Portrait4D without motion-related cross-attentions, trained with synthetic data of GenHead at 512x512 resolution
- portrait4d-v2-vfhq512: Portrait4D-v2 finetuned from portrait4d-static-genhead512, trained on VFHQ dataset at 512x512 resolution