Text-to-Speech on LibriSpeech clean
4CERWhisper + Llama 2 + MMS TTS
Evaluation Results
| Method | Links | |
|---|---|---|
| Whisper + Llama 2 + MMS TTSShots=10, Max context length=2048, Approach=Cascade Topline2024.02 | 4 | |
| SPIRIT LM BASE (+ASR+TTS)Shots=10, Max context length=20482024.02 | 6.7 | |
| ASR+TTS onlyShots=10, Max context length=2048, Training=Parallel Data2024.02 | 8.1 | |
| SPIRIT LM BASEShots=10, Max context length=20482024.02 | 45.5 | |
| SPIRIT LM EXPRESSIVEShots=10, Max context length=20482024.02 | 52 | |
| Word-level transcriptionShots=10, Max context length=2048, Training=Parallel Data2024.02 | 85.2 |