Dia
Ultra-realistic conversational TTS from Nari Labs — multi-speaker dialogue in one pass
Ultra-realistic conversational TTS from the two-person Nari Labs team; generates multi-character dialogue speech in one pass with full naturalness.
Voice AI: lifelike TTS, accurate ASR, and real-time voice agents.
Key features
- 1.6B parameter text-to-speech model from Nari Labs
- Generates realistic dialogue from a transcript with speaker tags
- Audio conditioning for emotion and tone control
- Nonverbal sounds: laughter, coughing, throat clearing and more
Good for
- Creators generating multi-speaker dialogue audio
- Developers building dialogue TTS into apps via Transformers
Stack & tags
PythonPyTorchTransformersGradiottsdialogueopen-weightsvoice-cloning1.6btransformers
Quick start
pip install git+https://github.com/nari-labs/dia.git
python example/simple.pyFrom the project README on GitHub.
Similar open-source projects
TEN-Agent
★ 11,144Open-source real-time voice-interaction agent platform — the "agent framework" for voice
NVIDIA PersonaPlex
★ 10,588NVIDIA's 7B full-duplex voice conversation model — customizable voices and personas