@MisoTTS

Miso Labs released MisoTTS, an open-weights 8-billion-parameter text-to-speech model that generates expressive speech from text and audio context. The model uses residual vector quantization to expand…