Linear correlation between human MOS and automatic metrics on TextrolSpeech (test)
0.749LCCSPAM (WavLM)
Evaluation Results
| Method | Links | |
|---|---|---|
| SPAM (WavLM)TTS Model=ParlerTTS-mini2026.01 | 0.749 | |
| RA-CLAPTTS Model=Ground truth audio2026.01 | 0.726 | |
| SPAM (WavLM)TTS Model=Ground truth audio2026.01 | 0.721 | |
| SPAM (CLAP)TTS Model=ParlerTTS-mini2026.01 | 0.71 | |
| RA-CLAPTTS Model=ParlerTTS-mini2026.01 | 0.709 | |
| SPAM (CLAP)TTS Model=Ground truth audio2026.01 | 0.673 | |
| SPAM (WavLM)TTS Model=PromptTTS2026.01 | 0.665 | |
| SPAM (CLAP)TTS Model=ParlerTTS-large2026.01 | 0.653 | |
| SPAM (WavLM)TTS Model=ParlerTTS-large2026.01 | 0.648 | |
| RA-CLAPTTS Model=ParlerTTS-large2026.01 | 0.615 | |
| SPAM (CLAP)TTS Model=PromptTTS2026.01 | 0.584 | |
| SPAM (CLAP)TTS Model=PromptStyle2026.01 | 0.518 | |
| SPAM (WavLM)TTS Model=PromptStyle2026.01 | 0.515 | |
| RA-CLAPTTS Model=PromptStyle2026.01 | 0.503 | |
| RA-CLAPTTS Model=VoxInstruct2026.01 | 0.473 | |
| SPAM (WavLM)TTS Model=VoxInstruct2026.01 | 0.462 | |
| SPAM (CLAP)TTS Model=VoxInstruct2026.01 | 0.429 | |
| RA-CLAPTTS Model=PromptTTS2026.01 | 0.416 |