We present Diff-TTSG, the first diffusion model that jointly learns to synthesise speech and gestures together. Our method is probabilistic and non-autoregressive, and can be trained on small datasets ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results