Text-to-Speech on LibriSpeech 2.2 hours subset (test-clean)
0.019WERVoicebox
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Voicebox#Param.=330M, #Data=60K EN2024.10 | 0.019 | 0.662 | 0.64 | |
| MELLE#Data=50K EN2024.10 | 0.021 | 0.625 | 0.538 | |
| MELLE-R2#Data=50K EN2024.10 | 0.0214 | 0.608 | 0.276 | |
| Ground TruthSubset=2.2 hours subset2024.10 | 0.022 | 0.754 | — | |
| VALL-E 2#Data=50K EN2024.10 | 0.0244 | 0.643 | 0.732 | |
| DiTTO-TTS#Param.=740M, #Data=55K EN2024.10 | 0.0256 | 0.627 | 0.162 |