Speech Reconstruction on English Read by Japanese accented speech 2007 (OOD)
14.9WERGround Truth
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Ground Truth2026.01 | 14.9 | 8 | 3.73 | — | — | — | |
| DualCodec 1:8trained on relevant data=true2026.01 | 17.1 | 9.4 | 3.71 | 0.98 | 0.86 | 0.07 | |
| X-Codec 2trained on relevant data=true2026.01 | 20.7 | 11.3 | 3.69 | 0.97 | 0.86 | 0.08 | |
| BiCodec2026.01 | 21.4 | 11.7 | 3.76 | 0.97 | 0.86 | 0.07 | |
| Kanade 25Hz2026.01 | 22.9 | 12.3 | 3.75 | 0.96 | 0.86 | 0.07 | |
| PAST 1:82026.01 | 25.3 | 14.1 | 3.65 | 0.97 | 0.85 | 0.07 | |
| Kanade 12.5Hz2026.01 | 33.8 | 18.6 | 3.78 | 0.95 | 0.8 | 0.09 | |
| StableCodec2026.01 | 51.4 | 29.3 | 4.03 | 0.91 | 0.87 | 0.06 | |
| WavTokenizer2026.01 | 51.7 | 31.6 | 3.06 | 0.91 | 0.82 | 0.08 |