Commit History
4 days ago
-
T2U v3 (12 decoder layers, trained with clone-v3 re-spoken mixed/en/fr data): CER 5.0% -> 4.2% on real voices; new demo
Verified signaturea376a72
6 days ago
-
Model card: Listen section with demo audio (Jarvis, zero-shot clones of synthetic voices), transcripts
Verified signature86b320c