Speaker Disentanglement on seed-tts-eval
2.03WERRaw audio
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Raw audio2025.07 | 2.03 | — | |
| ContentVec (ref.)Params=L122025.07 | 2.09 | 5.47 | |
| No filter2025.07 | 2.11 | 18.82 | |
| SHAP NoiseParams=σ = −0.32025.07 | 2.13 | 7.23 | |
| SHAP NoiseParams=σ = −0.62025.07 | 2.16 | 2.21 | |
| SHAP CroppingParams=α = 0.582025.07 | 2.18 | 10.07 | |
| SHAP CroppingParams=α = 0.992025.07 | 2.23 | 4.65 | |
| SHAP NoiseParams=σ = −1.02025.07 | 2.3 | 0 |