Dia
Nari LabsNari Labs' dialogue text-to-speech models, which turn a transcript tagged with two speakers into English conversation, including nonverbal sounds such as laughter and coughing. Dia2, in 1B and 2B sizes, streams: it starts speaking before the whole text arrives and can be conditioned on earlier audio, up to two minutes per generation. The first Dia, a 1.6B model, generates the dialogue in one pass.
Openness
3 high confidence- weights
- open (Dia2 1B and 2B checkpoints on the Hub, ungated)
- data
- closed (no training-data statement in the README or on the cards)
- code
- partial (inference code only)
- license
- Apache-2.0 (Dia2 weights and code)+CC-BY-4.0 (the Kyutai Mimi codec, which Dia2 downloads on first run)
- superseded-release
- Dia-1.6B (2025, Apache-2.0, still published)
Dia2 downloads freely under Apache-2.0, and the Kyutai audio codec it needs is CC-BY-4.0, which asks only for attribution. Nari Labs publishes inference code only and does not describe its training data.
- https://huggingface.co/api/models?author=nari-labs&limit=100 recorded 2026-11-12
Hub listing: Dia2-1B and Dia2-2B created 2025-11-15, Dia-1.6B 2025-04-20 and Dia-1.6B-0626 2025-06-26, all tagged "license:apache-2.0".
- https://huggingface.co/api/models/kyutai/mimi recorded 2026-11-12
Hub record for kyutai/mimi: "gated":false, cardData "license":"cc-by-4.0".
- https://huggingface.co/api/models/nari-labs/Dia2-2B recorded 2026-11-12
Hub record: "gated":false, cardData "license":"apache-2.0".
- https://raw.githubusercontent.com/nari-labs/dia/HEAD/LICENSE recorded 2026-11-12
Dia LICENSE body: "Apache License" / "Version 2.0, January 2004".
- https://raw.githubusercontent.com/nari-labs/dia2/HEAD/LICENSE recorded 2026-11-12
LICENSE body: "Apache License" / "Version 2.0, January 2004".
- https://raw.githubusercontent.com/nari-labs/dia2/HEAD/README.md recorded 2026-11-12
README: "We provide model checkpoints (1B, 2B) and inference code to accelerate research."; "The first run downloads weights/tokenizer/Mimi."; "Licensed under [Apache 2.0](LICENSE). All third-party assets (Kyutai Mimi codec, etc.) retain their original licenses."
Adoption
2 high confidenceHugging Face downloads over the trailing month, summed across the two Dia2 and two Dia 1.6B checkpoints, most of them for the June 2025 Dia 1.6B update.
- https://huggingface.co/api/models?author=nari-labs&limit=100 recorded 2026-11-12
"downloads": Dia-1.6B 5557, Dia-1.6B-0626 21286, Dia2-1B 1799, Dia2-2B 7430.
Capability
1 medium confidenceDia has no listener rating and publishes no results on a public test set, so nothing places it beside the rated synthesizers. It sits with MeloTTS, another open voice with no published evaluation.
- https://artificialanalysis.ai/text-to-speech/leaderboard recorded 2026-11-12
The 104-row leaderboard has no Dia entry.
- https://raw.githubusercontent.com/nari-labs/dia2/HEAD/README.md recorded 2026-11-12
README: "**Dia2** is a **streaming dialogue TTS model**"; "The model only supports up to 2 minutes of generation in English."; no evaluation section.
Verified 2026-11-12