Multi-speaker Transcription on AISHELL-4 Chinese
6.77DERVIBEVOICE-ASR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VIBEVOICE-ASRInference Protocol=Single-pass2026.01 | 6.77 | 24.99 | 25.35 | 21.4 | |
| Gemini-2.5-ProInference Protocol=Chunk-wise, Chunk Duration=240s2026.01 | 15.32 | 31.59 | 35.96 | 22.42 | |
| Gemini-3-ProInference Protocol=Chunk-wise, Chunk Duration=240s2026.01 | 22.03 | 27.43 | 54.17 | 22.75 |