AURA
A Kyutai family for a talking avatar driven by sound: from a portrait and a speech stream it draws a talking head piece by piece with a rolling cache. The 1.3B model is the only one in the family that runs in real time: 576p and, by the authors paper, 20.7 frames per second end to end on a single H100, 23 times faster than its teacher; it is a four-step causal student. A larger AURA-5B exists too. The license is non-commercial, and it cannot be used in paid products.
- Developer
- Kyutai, France
- First release
- Nov 2026
- Latest release
- Nov 2026
- Sizes
- 1.3B and 5B
- License
- Non-commercial onlyCC BY-NC 4.0: non-commercial use only
- Running
- On your own serverNeeds a GPU
- Industries
- Media and production, Customer support
What it does
- research and prototypes of live avatars in real time
- comparison with other talking-head models
- personal and educational projects without commerce
Where it is used
Hardware requirements
Versions
- AURA-5B
- AURA-1.3B (реальное время)
How to run it
I can set this up end to end: pick the model size, deploy it on your server and connect it to your systems.
Frequently asked questions
Can AURA be used in a commercial project?
No. License: CC BY-NC 4.0: non-commercial use only. A commercial product needs a different model or a separate agreement with the rights holder.
What hardware does AURA need?
At minimum: One GPU with 16–80 GB — mid-size versions. Without a GPU the model is not practical. You can calculate the exact VRAM for your model size and context in the hardware calculator.
Does AURA support Russian?
Language does not matter for this model: it does not work with text.
Where can I download AURA and what does it cost?
The AURA weights are open and free to download. You only pay for the hardware it runs on and for the setup. Source links are at the bottom of this page.
How I deploy it for clients
- SelectionI pick the model size for your task and hardware and test it on your examples.
- DeploymentI deploy it on your server or in a closed network and provide an API.
- Fine-tuningI fine-tune it on your data (LoRA) or connect a knowledge base — whichever is cheaper for the task.
- IntegrationI connect it to your CRM, ERP, bot, website or team chat and set up monitoring.
Similar models
Audio-driven talking people built on LongCat-Video. Version 1.5 is production-ready: stable long videos in Chinese and English. Version 1.6 (November 13, MIT) gives more natural and stable lip animation, coordinated face and body, better consistency in long videos and faster deployment.
DetailsAvatarsHunyuanVideo-AvatarTencent · ChinaCommercial use with conditionsTalking characters built on HunyuanVideo: conveys emotions from the voice, handles several characters and different styles.
DetailsVoice assistantsMoshi / HibikiKyutai · FranceCommercial use with conditionsA voice assistant that listens and speaks at the same time, with no delay for recognition and synthesis. Hibiki does simultaneous speech-to-speech translation between several European languages.
DetailsSource: huggingface.co/kyutai/AURA-1.3B. Data checked against the model card on 22 Sep 2026. Have a lawyer review the license before commercial launch.


