nari-labs/dia2
A streaming dialogue text-to-speech model that generates audio in real-time as text is provided.
Not currently ranked — collecting fresh signals.
star history
Dia2 is an open-weight TTS model from Nari Labs with 1B and 2B parameter variants capable of real-time streaming audio generation. The model can begin producing speech as the first few words are provided, and supports conditioning on audio input to enable natural conversational dialogues. Inference is available via a CLI that supports CUDA graph optimization and bfloat16 precision, with model weights hosted on Hugging Face.
Frequently asked
- What is nari-labs/dia2?
- A streaming dialogue text-to-speech model that generates audio in real-time as text is provided.
- Is dia2 open source?
- Yes — nari-labs/dia2 is open source, released under the Apache-2.0 license.
- What language is dia2 written in?
- nari-labs/dia2 is primarily written in Python.
- How popular is dia2?
- nari-labs/dia2 has 1.2k stars on GitHub.
- Where can I find dia2?
- nari-labs/dia2 is on GitHub at https://github.com/nari-labs/dia2.