Environment-aware Text-to-Speech on LibriTTS and AudioCaps (test)
2.21WERGround Truth
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Ground Truth2026.05 | 2.21 | 4 | 92.18 | — | 51.2 | |
| ImmersiveTTSNFEs=252026.05 | 9.89 | 3.23 | 88.59 | 7.81 | 32.3 | |
| VoiceDiTNFEs=2002026.05 | 11.08 | 3.33 | 89.42 | 8.23 | 30.2 | |
| VoiceLDMNFEs=2002026.05 | 14.01 | 2.82 | 76.01 | 8.71 | 28.8 |