preprint · arXiv (2025)
DeepGesture: A conversational gesture synthesis system based on emotions and semantics
Thanh Hoang-Minh
In plain English
A diffusion model that generates co-speech gestures from text, speech, emotion and seed motion, with controllable emotional style. Evaluated on ZeroEGGS, it improves human-likeness and generalizes to synthetic voices.
gesture generation
digital humans
diffusion
audio-driven animation