Singing Voice Style Transfer on Multi-singer unseen (M4Singer, OpenSinger, PopBuTFy, GTSinger) (test)
4.58MOS-QGT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GTType=Ground Truth2025.05 | 4.58 | — | — | — | |
| GT (vocoder)Vocoder=HiFi-GAN2025.05 | 4.36 | 4.41 | 0.04 | 0.95 | |
| TCSinger 2Type=Proposed Flow-matching SVS2025.05 | 4.13 | 4.27 | 0.21 | 0.93 | |
| TCSinger 2Type=Proposed Flow-matching SVS2025.05 | 3.96 | 4.09 | — | — | |
| TCSingerType=Zero-shot SVS2025.05 | 3.94 | 4.01 | 0.26 | 0.91 | |
| VISinger 2Type=High-fidelity SVS2025.05 | 3.79 | 3.88 | 0.31 | 0.83 | |
| TCSingerType=Zero-shot SVS2025.05 | 3.77 | 3.87 | — | — | |
| CosyVoiceConfiguration=Enhanced with note encoder2025.05 | 3.74 | 3.93 | 0.33 | 0.87 | |
| StyleTTS 2Configuration=Enhanced with note encoder2025.05 | 3.71 | 3.79 | 0.42 | 0.71 | |
| VISinger 2Type=High-fidelity SVS2025.05 | 3.69 | 3.72 | — | — | |
| CosyVoiceConfiguration=Enhanced with note encoder2025.05 | 3.63 | 3.77 | — | — | |
| StyleTTS 2Configuration=Enhanced with note encoder2025.05 | 3.58 | 3.63 | — | — |