Remove background

One tap to remove the background. Fine-tune if needed.

Scroll down for tips, limits, and FAQs

Automatic cutouts and manual edge work

Pillar guide: How to Remove Backgrounds for Product Photos

Marketplace listings fail review more often for halos and fringe around bottles and packaging than for slightly soft interiors. Start with automatic cutout when contrast is good, then brush edges at 200–400% zoom.

Automatic removal runs on-device in your browser (model weights download once; your photo is not sent to our servers for inference). Manual brush and polygon tools finish hair, glass, and reflective rims.

Validate exports on both white (#FFFFFF) and dark backgrounds before publishing—checkerboard previews hide light halos that show up on storefront templates.

Background removal isolates a subject so you can place it on white, transparent, or custom scenes. Start with the purple “Remove background” button for an automatic AI cutout, then refine edges with the brush or polygon tools when needed.

The automatic mode runs an on-device segmentation model in your browser (via rembg-webgpu, using WebGPU when available or WASM as a fallback). Your image is not uploaded to our servers for that step—inference happens locally after the model is ready.

On the first visit, the browser may download model weights once and cache them; later runs are much faster. After the automatic mask is applied, zoom in on hair, glass, and product edges and touch up with manual tools before exporting PNG or WebP with transparency.

Automatic and manual workflow

Upload a photo; the editor prepares a working canvas sized for your device.
Click “Remove background” (One-tap cutout) to run automatic AI removal. The first run can take longer while the model initializes; progress appears in the tool.
Review the result on the checkerboard preview. Use Brush to erase or restore areas, or Polygon to define a precise shape when the AI mask is close but not perfect.
Undo and redo mask edits as needed, then export transparent PNG or WebP.

Common use cases

Quick cutouts for social posts, slides, and mockups when subject and background contrast well.
Marketplace listings that require white or transparent backgrounds.
Product and portrait workflows where you want AI speed plus manual edge control.

Edge cleanup tips

Use automatic removal first on clean subjects, then brush only where the mask fringes or misses fine detail.
Subjects with clear separation from the backdrop usually get the best automatic results.
On phones, inference may run on a slightly smaller working size to keep memory stable—refine critical edges at zoom afterward.
Validate exports on both white and dark backgrounds to catch halos before publishing.

Limitations to know

Automatic AI quality depends on lighting, clutter, and subject type; complex fur, motion blur, or reflections often need manual brush work.
The first automatic run requires a one-time model download in the browser (network needed for weights, not for sending your photo to us).
Very old browsers without WebGPU/WASM support may fail automatic mode; manual brush and polygon still work once the editor loads.

Related guides

Frequently asked questions

What does the automatic Remove background button do?

It runs a local AI segmentation model that estimates which pixels belong to the subject and builds an alpha mask. You can then refine that mask with brush or polygon tools and export a transparent file.

Is my image sent to a server for AI removal?

No. Automatic cutout processing is designed to run in your browser. Only the initial model download (first visit) uses the network; your photo stays on your device during inference.

Why is the first automatic run slower?

The browser downloads and caches the model weights on first use. Later automatic runs in the same browser are typically much faster.

Which export format keeps transparency?

PNG and WebP support alpha channels; JPEG does not.