EmoVOCA: Speech-Driven Emotional 3D Talking Heads

arXiv:2403.12886v1 Announce Type: new
Abstract: The domain of 3D talking head generation has witnessed significant progress in recent years. A notable challenge in this field consists in blending speech-related motions with expression dynamics, which is primarily caused by the lack of comprehensive 3D datasets that combine diversity in spoken sentences with a variety of facial expressions. Whereas literature works attempted to exploit 2D video data and parametric 3D models as a workaround, these still show limitations when jointly modeling the …

abstract arxiv challenge cs.cv datasets diversity domain dynamics head progress speech spoken talking head type