Zero-shot Text-to-Speech on LibriSpeech SNR = ∞ (test-clean)
3.76UTMOSF5-TTS
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| F5-TTSTraining Dataset=LibriTTS2025.05 | 3.76 | 24 | 53 | — | 80 | 13.78 | 67 | 0.01 | |
| VALL-ETraining Dataset=LibriTTS2025.05 | 3.68 | 19 | 40 | 48 | 75 | 21.66 | 36 | 0.02 | |
| VoiceCraftTraining Dataset=GigaSpeech2025.05 | 3.55 | 18 | 51 | 45 | 78 | 17.22 | 44 | 0.01 | |
| OZSpeechTraining Dataset=LibriTTS, Fine-tuned on noisy prompts=true2025.05 | 3.19 | 6 | 39 | 46 | 78 | 13.67 | 65 | 0.01 | |
| OZSpeechTraining Dataset=LibriTTS2025.05 | 3.15 | 5 | 39 | 47 | 81 | 11.96 | 67 | 0.01 | |
| NaturalSpeech 2Training Dataset=LibriTTS2025.05 | 2.38 | 9 | 31 | 38 | 80 | 15.62 | 25 | 0.02 |