withoutbg/withoutbg-openweights-v4-app-gpu repository overview

⁠withoutBG Open Weights v4 — Web App (GPU)

Self-hosted withoutBG⁠ background removal web app with GPU-accelerated inference. Upload images in the browser and remove backgrounds using the open weights v4 ONNX model on CUDA.

New in v4: Smart brush. Paint what to keep and what to remove; a small stroke-edit network fixes the cutout. Open a result and choose Smart brush.

⁠Quick start

Requires an NVIDIA GPU and the NVIDIA Container Toolkit⁠.

docker run --rm --gpus all -p 8080:8080 withoutbg/withoutbg-openweights-v4-app-gpu:latest

Open http://localhost:8080 in your browser.

The bundled API is available at http://localhost:8080/api (proxied to the internal FastAPI service).

⁠Platforms

Published for linux/amd64 only (NVIDIA CUDA on x86_64).

⁠What's included

• Drag-and-drop web UI for background removal
• FastAPI inference service (/api/v1/remove-background)
• Health endpoint at /health

⁠Optional: withoutBG Pro

Set -e WITHOUTBG_API_KEY=sk_your_key -e WITHOUTBG_PRO_PASSWORD=<long random password> and an Open weights / Pro switch appears in the web UI. Pro uploads the image to api.withoutbg.com (1 credit each). The open weights stay the default and never send images anywhere. Pro calls are charged to your key, so each one must send the password (X-Pro-Password; the web UI asks once). Keep a Pro-enabled server private, e.g. -p 127.0.0.1:8080:8080. Keys: withoutBG Pro⁠.

⁠Model

The withoutBG Open Weights 10.8.0 bundle (~1.5 GB, baked into the image): a trained router sends fine strands, soft detail and transparency to the withoutBG matting model, and hard opaque objects, flat scenes and vehicles to BiRefNet. Only the selected branch runs.

⁠Related images

⁠Links

• withoutBG open model⁠
• Source on GitHub⁠
• License: Apache-2.0 for the code; the model is under the withoutBG Open Weights license⁠ and includes DINOv3, Depth Anything V2 (Apache-2.0) and BiRefNet (MIT). See /v1/licenses in the running container