List your company

Continuation Zero-shot Speech Synthesis on LibriSpeech (test-clean)

1.98WERHMELLE
1.90722.39862.893.3814Jul 11, 2024

Evaluation results

MethodLinks
MELLETraining Data Hours=50K2024.071.981.470.508
VoiceboxTraining Data Hours=60K2024.072—0.593
MELLE-R2Training Data Hours=50K, reduction factor=22024.072.021.450.489
MELLE-R3Training Data Hours=50K, reduction factor=32024.072.11.520.462
MELLE-R4Training Data Hours=50K, reduction factor=42024.072.11.590.437
Ground Truth2024.072.151.610.668
MELLE-limitedTraining Data Hours=9602024.072.221.530.48
Ground Truth (mel-spectrogram)2024.072.241.640.617
MELLE-R5Training Data Hours=50K, reduction factor=52024.072.251.660.41
VALL-E RTraining Data Hours=9602024.072.321.580.363
VALL-E 2Training Data Hours=50K2024.072.321.60.504
Ground Truth (EnCodec, 8 codebooks)2024.072.331.650.593
CLaM-TTSTraining Data Hours=55K2024.072.36—0.477
ELLA-VTraining Data Hours=9602024.072.912.10.303
VALL-ETraining Data Hours=60K2024.073.8—0.508