SynchroRaMa : Lip-Synchronized and Emotion-Aware Talking Face Generation via Multi-Modal Emotion Embedding #57
Description
https://arxiv.org/pdf/2509.19965 -
the losses look good -
SyncLoss. - Emo Loss. Facial Action Unit (AU) Loss. - my latest dataset has the VA extracting - maybe can cherry pick this loss function when code drops.