List your company

Voice Conversion on ELHE HE portion (160 unmodified utterances)

2.9CERGT
-2.59634.50271.6108.698Jan 7, 2026

Evaluation results

MethodLinks
GTDescription=Ground-truth healthy speech2026.012.943.484.113.20.89—
Ours (adapted StreamVC)Guided loss=WavLM2026.0140.93.263.323.932.940.840.34
Ours (adapted StreamVC)Guided loss=WavLM + HF2026.0141.93.763.4343.090.870.34
Ours (adapted StreamVC)Guided loss=WEO + HF2026.0144.93.693.394.023.050.860.34
Ours (adapted StreamVC)Guided loss=WavLM + HF + F02026.0146.73.73.423.973.060.870.35
Ours (adapted StreamVC)Guided loss=WEO + WavLM + HF2026.0147.13.73.383.983.020.870.34
Ours (adapted StreamVC)Guided loss=none2026.0153.73.173.293.882.90.860.35
Ours (adapted StreamVC)Guided loss=BNF + HF2026.0155.413.823.423.983.070.880.35
XVCModel type=Offline SOTA VC2026.0161.23.593.324.0230.630.37
ELDescription=Electrolaryngeal speech source2026.0188.2-0.283.143.122.410.550.62
FreeVCModel type=Offline SOTA VC2026.01140.33.523.273.992.910.710.4