One command installs a licensed model on your GPU and runs it:
npx @bria-ai/studioStudio signs you in through your browser, installs the model you pick into a virtual environment at ~/.bria/venvs/<model>, downloads its weights, and runs a first generation.
Before you run it
You need Node.js, an NVIDIA GPU, and a model license. No license yet? See pricing and Self-hosted models.
The model's package stays installed in that virtual environment. Activate it and call the model from your own code:
source ~/.bria/venvs/rmbg-2.0/bin/activatefrom remove_bg import RemoveBG, RemoveBGConfig, RemoveBGInput
pipe = RemoveBG(config=RemoveBGConfig(api_token="<your-bria-api-token>"))
pipe.setup() # reuses the weights Studio already downloaded
pipe.execute([RemoveBGInput(image="input.jpg")])[0].image.save("cutout.png")- Full options, and installing without Studio — each model's guide: Fibo 2, RMBG 2.0
- Running in production? Self-hosting best practices
- How licensing, API keys and usage work: Self-hosted models