# nari-labs/dia

A TTS model capable of generating ultra-realistic dialogue in one pass.

Repository: https://github.com/nari-labs/dia
Canonical: https://ross.abutalabs.com/products/dia
Language: Python
License: Apache-2.0
License Family: permissive
Topics: ai, open-weight, text-to-speech
Last push: 2025-11-19T21:11:55+00:00

## Health v2 (maintenance only)
Score: 43/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 53, release rhythm 35, longevity 35
- inputs: {"age_days": 501, "days_push": 287, "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 19378, forks 1689 (observed 2026-08-28T04:11:27.673260+00:00)

## What it is
Dia is a 1.6B-parameter open-weight text-to-speech model from Nari Labs that generates ultra-realistic multi-speaker dialogue in a single pass, with conditioning on audio for emotion and tone control. It ships with inference code and pretrained checkpoints on Hugging Face, and supports nonverbal cues like laughter.

## Use cases
- generate realistic two-speaker dialogue audio from a script
- clone a voice from an audio prompt for TTS
- add nonverbal sounds like laughter to synthesized speech
- control emotion and tone of generated speech
- build a podcast or audiobook generator
- run a local text-to-speech model with open weights

## When to choose
- you need realistic conversational speech, not single-voice narration
- you want an open-weight TTS model you can self-host
- you need speaker turns and nonverbal cues in one generation

## When to avoid
- you need multilingual TTS - it only supports English
- you need very short or very long utterances, which degrade quality
- you lack a GPU for inference

## Facets
- artifact type: library
- maturity: active
- function: tts, machine-learning, llm-inference
- domain: speech-processing, artificial-intelligence, deep-learning
- platform: python, cross-platform
- tags: text-to-speech, open-weights, dialogue-generation, voice-cloning, huggingface, audio, gpu

## Member repositories
- nari-labs/dia (main) score 43

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:27.673260+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 2026-08-29T17:00:09.786496+00:00, confidence not recorded.
  - readme: https://github.com/nari-labs/dia (fetched 2026-08-28T04:11:27.673260+00:00, sha 37996b111669)
- Data as of 2026-08-30T08:39:29.467469+00:00.