A self-hosted license lets you run a Bria model on your own GPUs instead of calling Bria's cloud. You buy one license per model, install the model's package, and run it from your own services. Your images and video never leave your environment, and the weights come from Bria directly — you never need a Hugging Face account.
Each model is licensed on its own. Buy a license from the pricing page, then manage it on the Licenses tab in the Bria Console. Companies with revenue under $10M can buy Starter and Scale by card; larger companies and custom needs go through Bria sales.
- License. Buy a per-model license in the Console. A license is a subscription in your billing account — nothing is installed by buying it.
- Install. On your own machine, install the model's package and download its weights. The fastest way is Bria Studio; you can also install the Python library directly.
- Run. Load the model with one of your API keys and run it. Every generation is counted against your license and shown in the Console.
| Model | What it does | Guide |
|---|---|---|
| Fibo 2 | Image generation and editing | Fibo 2 |
| RMBG 2.0 | Background removal for images | RMBG 2.0 |
More models are on the way. The pricing page shows the full launch list and marks which are available today.
Two ways to install a licensed model — Bria Studio, or a manual install with one of your API keys:
Installs the model, downloads the weights, and runs a test. Needs Node.js and a GPU.
npx @bria-ai/studioStudio signs you in through your browser (no token to copy) and lists your licensed models. See the quickstart.
- Any of your API keys works. No separate license key — manage keys on the API keys page and set one as
BRIA_API_TOKEN. - Usage shows in the Console. Self-hosted runs appear in Usage & billing, tagged as self-hosted.
- A lapsed license stops new runs. A cancelled or ended license refuses new starts; a running process continues until its next restart. A card retry (
past_due) stays active. - Buying a license never blocks the API. License and API plan are separate. With no API plan, a license moves you to pay-as-you-go; an existing plan is unchanged.
Self-hosted licenses run the published model version on your own cloud GPUs. For air-gapped or on-premises deployments, full-precision models, automatic upgrades, an SLA, or custom models, see Deployment options or contact sales.