List your company

Text-to-Speech Generation on LibriSpeech-PC (test-clean)

1.93WERTAS-VAE
1.81082.61543.424.2246Jun 30, 2025Aug 17, 2025Oct 4, 2025Nov 21, 2025Jan 8, 2026Feb 25, 2026Apr 14, 2026

Evaluation results

MethodLinks
TAS-VAEAdaptive Weighting=true2026.041.933164.5
Semantic-VAEAdaptive Weighting=false2026.042.016782.5
JMAS-VAEAdaptive Weighting=true2026.042.045777.5
Baseline (Mel/Fbank)Adaptive Weighting=false2026.042.236179.4
MegaTTS 32025.062.3170—
DAS-VAEAdaptive Weighting=true2026.042.323264.8
F5-TTS2025.062.4266—
DAS-VAEAdaptive Weighting=false2026.042.465978.3
TAS-VAEAdaptive Weighting=false2026.042.556380.2
JMAS-VAEAdaptive Weighting=false2026.042.596380.2
FireRedTTS2025.062.6947—
Vanilla VAEAdaptive Weighting=false2026.042.725877.6
E2 TTS2025.062.9569—
Cosy Voice2025.063.5966—
EnCodecAdaptive Weighting=false2026.044.715675.6
JAM-Flowaudio-only=true2025.064.9164—