MISO-TTS: 8B parameter text-to-speech model released

View on Reddit

Miso Labs released MISO-TTS, an 8 billion parameter text-to-speech model based on the Sesame CSM architecture. It uses a Llama 3.2-style backbone to generate Mimi audio codes from text and optional audio context.