Image🔒 Runs in your browser

Background Remover

Remove the background from any image in one click

+Drop an image here or click to browse

Use cases

Drop a product photo onto a clean white background for an e-commerce listing.
Isolate a portrait for a profile picture with a transparent edge.
Extract a logo from a flat-coloured photo to add it as an overlay.
Prepare a person cutout to composite onto a different scene in a slide.
Remove a busy background from a phone selfie before adding it to a CV.

About Background Remover

Image Background Remover erases the background of any photo with a one-click ML pipeline that runs entirely in your browser. The U²-Net image segmentation model (4–24 MB depending on quality) loads on first use via ONNX Runtime Web, then segments subject from background per-pixel — no upload, no server. Use it to drop a product photo onto a clean white background for a shop listing, isolate a portrait for a profile picture or social card, prepare a transparent logo from a flat-coloured photo, or reframe a person against a different scene. The result is a transparent PNG you can layer onto any background.

Frequently Asked Questions

Is my image uploaded to a server?

No. The AI model runs entirely in your browser using ONNX Runtime Web with WebAssembly. Your image never leaves your device.

Which file formats are supported?

You can drop any common image format (PNG, JPG, WebP, GIF, BMP). The export is either PNG (for transparency) or WebP (smaller file size, adjustable quality).

How accurate is the background removal?

The underlying AI model is tuned for photos with a clearly defined subject (people, products, objects). You can fine-tune the edge smoothness with the Feather slider to blend hair, fur or fine details more naturally.

Why is the first run slow?

The first time you remove a background, the browser downloads the ONNX model (4 MB for the lightweight U²-Net-Light, up to 24 MB for the full U²-Net) and the ONNX runtime WASM (~13 MB). After this one-time download, subsequent runs use the cached model and are ~5× faster. The Cache-Control headers are set to keep these assets for one year.

How accurate is the segmentation?

Excellent on portraits, products on plain backgrounds, and animals against contrasting scenes. Hair edges and translucent fabrics (lace, mesh) can produce slight artefacts. The full U²-Net model handles edge cases better than the Light variant — the tool auto-picks based on connection speed but you can override.

🔧 Related Tools

Use cases

Drop a product photo onto a clean white background for an e-commerce listing.

Isolate a portrait for a profile picture with a transparent edge.

Extract a logo from a flat-coloured photo to add it as an overlay.

Prepare a person cutout to composite onto a different scene in a slide.

Remove a busy background from a phone selfie before adding it to a CV.