Miso One is the new Miso Labs voice model: an 8B open-weights text-to-speech system built for expressive English conversational speech, voice continuation, and low-latency voice-agent research. It supports one-shot voice cloning, local deployment, and real-time voiceover workflows with 48 kHz audio previews.
Key benefits include:
- Expressive English Conversational Speech: Delivers natural, emotive speech with emotional variation, pauses, and conversational rhythm for creator workflows.
- Low-Latency Generation: Achieves a claimed 110 ms latency, ideal for interactive voice-agent responses and real-time voiceover tasks.
- Open-Weight Access: Public MisoTTS repository and Hugging Face weights enable local inference, development, and customization for developers.
- Voice Cloning & Continuation: Supports one-shot voice cloning via audio context, allowing voice identity preservation and continuation for projects.
- Local Deployment: Developers can run the 8B checkpoint in their own CUDA environments, with hardware requirements and setup guidance provided.
Perfect for developers, voice-agent researchers, and content creators seeking expressive English conversational speech with low-latency, open-weights text-to-speech for local or agent workflows.
