Moshi
Speech-text foundation model for full-duplex real-time spoken dialogue with neural audio codec.
About
Moshi is a speech-text foundation model from Kyutai for full-duplex real-time spoken dialogue. It processes two simultaneous audio streams while predicting text tokens, and bundles Mimi, a streaming neural audio codec running at 12.5 Hz with 1.1 kbps. The repository ships PyTorch, MLX, and Rust backends covering research, Apple Silicon, and production deployments.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- Text-to-Speech (TTS)
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Advanced (4/5)
- License
- Apache-2.0
- Minimum VRAM
- 24 GB
- Added
- May 7, 2026
Related Tools
Lightweight and expressive TTS model with 82M parameters for fast local inference.
Open-source TTS model by Resemble AI with emotion and accent control.
Expressive zero-shot TTS model by Resemble AI with emotion and accent control.
Singing voice conversion model based on VITS and SoftVC for voice-to-voice transfer.
Zero-shot TTS model with high naturalness and speaker similarity.
Transformer-based text-to-audio model by Suno that generates speech, music, and sound effects.