Audiovisual Synchronization on Seamless Interaction Free-run
10.1LSE-DMoshi-Face w/o t−1 face token input
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Moshi-Face w/o t−1 face token inputStreaming=true2026.06 | 10.1 | 9 | |
| Moshi-Face w/o Face Transformer pre-trainingStreaming=true2026.06 | 10.4 | 14 | |
| Moshi-FaceStreaming=true2026.06 | 11 | 16 | |
| Moshi-Face w/o full fine-tuningStreaming=true2026.06 | 11.1 | 20 | |
| Random faceStreaming=false2026.06 | 11.8 | 11 |