[SIGGRAPH Asia 2026] UniMate: One Unified Model to Animate Diverse Skeletons
"""Text encoders (T5, CLIP, BERT) behind a unified tokenize/forward interface."""