# nari-labs/dia2

TTS model capable of streaming conversational audio in realtime.

Repository: https://github.com/nari-labs/dia2
Canonical: https://ross.abutalabs.com/products/dia2
Language: Python
License: Apache-2.0
License Family: permissive
Topics: open-weight, text-to-speech
Last push: 2025-11-29T00:51:56+00:00

## Health v2 (maintenance only)
Score: 41/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 54, release rhythm 35, longevity 20
- inputs: {"age_days": 289, "days_push": 278, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1172, forks 100 (observed 2026-08-28T04:03:51.698517+00:00)

## What it is
Dia2 is a streaming dialogue text-to-speech model from Nari Labs that can generate conversational audio in realtime without needing the full input text. It ships with 1B and 2B open-weight checkpoints, a Python inference library, a CLI, and a Gradio app, and supports conditioning output on audio prefixes for stable voices.

## Use cases
- generate speech from text in python
- stream conversational tts in realtime
- build a speech-to-speech voice assistant
- clone or condition a voice with audio prefixes
- generate multi-speaker dialogue audio with speaker tags
- run a local text-to-speech model on gpu

## When to choose
- you need realtime or streaming conversational TTS with open weights
- you want multi-speaker dialogue generation with [S1]/[S2] speaker tags
- you want to condition TTS output on prior audio for speech-to-speech pipelines
- you can run CUDA 12.8+ on a local GPU

## When to avoid
- you need stable, consistent voices without conditioning or fine-tuning
- you need languages other than English or generations longer than 2 minutes
- you need a CPU-only production TTS solution
- you need a fully managed or hosted TTS API

## Facets
- artifact type: library
- maturity: active
- function: tts, speech-recognition, llm-inference, audio-processing, cli
- domain: speech-processing, artificial-intelligence
- platform: python, cli, cross-platform
- tags: streaming-tts, dialogue-tts, voice-cloning, open-weights, conversational-audio, cuda, gradio, audio, natural-language-processing, gpu, linux

## Member repositories
- nari-labs/dia2 (main) score 41

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:51.698517+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:28:28.039862+00:00, confidence not recorded.
  - readme: https://github.com/nari-labs/dia2 (fetched 2026-08-28T04:03:51.698517+00:00, sha 6027eecd0b88)
- Data as of 2026-08-30T08:39:29.467469+00:00.