Zero-shot Text-to-Speech Synthesis on LibriSpeech 3s audio prompt (test-clean)
0.26RTFOZSpeech
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OZSpeechTrainable Parameters=145M, Frozen Parameters=102M, Codec=FACodec2025.05 | 0.26 | 1 | 5 | 40 | |
| F5-TTSTrainable Parameters=336M, Frozen Parameters=13.5M, Vocoder=Vocos2025.05 | 0.7 | 32 | 24 | 53 | |
| VALL-ETrainable Parameters=594M, Frozen Parameters=104M, Tokenizer=SpeechTokenizer2025.05 | 0.86 | — | 19 | 40 | |
| NaturalSpeech 2Trainable Parameters=378M, Frozen Parameters=14M, Codec=EnCodec2025.05 | 1.66 | 200 | 9 | 31 | |
| VoiceCraftTrainable Parameters=830M, Frozen Parameters=14M, Codec=EnCodec2025.05 | 1.7 | — | 18 | 51 |