Gesture Synthesis on BEAT2 multi-speaker (23 speakers)
0.684BeatAlignGestureLSM (Causal*)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GestureLSM (Causal*)Causal=true, Retrained=true2026.03 | 0.684 | 2.792 | 9.11 | — | |
| RAG-Gesture2026.03 | 0.648 | 0.515 | 10.09 | — | |
| GestureLSM2026.03 | 0.481 | 0.537 | 8.41 | — | |
| MIBURI(+Face)Model scale=Standard, Face modeling=true2026.03 | 0.461 | 0.48 | 10.44 | 7.77 | |
| GT2026.03 | 0.446 | — | 8.45 | — | |
| MIBURI-L (+Face)Model scale=Large, Face modeling=true2026.03 | 0.434 | 0.582 | 9.31 | 7.63 | |
| MIBURI-LModel scale=Large2026.03 | 0.431 | 0.555 | 9.45 | — | |
| MIBURIModel scale=Standard2026.03 | 0.415 | 0.585 | 9.75 | — | |
| EMAGE*Retrained=true2026.03 | 0.236 | 0.85 | 6.58 | 4.6 | |
| CaMN2026.03 | 0.176 | 0.736 | 6.73 | — | |
| MambaTalk (Causal*)Causal=true, Retrained=true2026.03 | 0.102 | 1.222 | 4.61 | 4.17 | |
| MambaTalk*Retrained=true2026.03 | 0.08 | 1.375 | 3.73 | 4.12 |