1 paper · 1 filter
Yihuan Huang, Jiajun Liu, Yanzhen Ren +3
Recent talking head synthesis works typically adopt speech features extracted from large-scale pre-trained acoustic models. However, the intrinsic many-to-many relationship between…