toolgarden.xyz
中文

Remove image background

Free online background remover that processes images locally in your browser with an open-source model and exports a transparent PNG

Background removal

Local model

Export transparent PNG

PNG

Use an open-source image segmentation model in the browser to detect the subject and remove the background. Balanced mode is the default, with a lighter fast mode and a more detailed HD mode available. Model assets are usually cached by the browser.

Model
Supported formats
JPGPNGWebPGIFBMPSVGAVIF
Current file--
Source size--

All file processing happens locally. Files are not uploaded to any server, and data is cleared when you close the page.

Background removal result

Waiting for an image

Upload one image, then remove the background. The transparent PNG result will appear on the right.

About this tool

Background removal identifies the main foreground and creates a transparency mask so a person, product, or object can be placed over another scene. The tool offers quality-oriented and speed-oriented models and processes in the browser; the selected model may need to load on first use.

Separation quality depends on boundary clarity, foreground-background contrast, and source resolution. Hair, transparent materials, shadows, fine mesh, and similarly colored regions are hardest, so inspect edges on a checkerboard and on both light and dark backgrounds before export.

How to use it

  1. Choose a clear source

    Prefer an evenly lit image with the whole subject visible and strong separation from the background.

  2. Select a model

    Use the faster model for routine previews and try the quality model for hair or detailed product edges.

  3. Inspect transparent edges

    Zoom into the outline, internal gaps, and translucent regions, then export a format with alpha support.

Supported range and limits

Output
A PNG with a transparent background
Model
An open-source segmentation model running in-browser; the first run downloads tens of megabytes
Works well on
Portraits, product shots and single-subject photos where subject and background contrast clearly
Common artefacts
Hair strands, fur, semi-transparent materials (glass, mesh, smoke) and edges the same colour as the background
Time required
Inference runs locally, so large images and modest hardware are noticeably slower
Next step
If you need a solid backdrop or a passport-sized crop, ID Photo Maker does both in one pass

When you would use it

  • Creating product artwork

    Isolate a product and place it on a brand color, lifestyle scene, or consistent store background.

  • Preparing a profile portrait

    Remove a distracting environment so the person can be used in a résumé, presentation, or social profile.

  • Re-backgrounding product photos

    Store listings need a consistent flat or styled backdrop, so the original background comes off first and the subject is composited onto the target.

What to know before you start

  • Automatic segmentation is an estimate and is not suitable as pixel-accurate evidence for legal, medical, or forensic work.
  • Save transparent output as PNG or alpha-capable WebP. JPG discards transparency.
  • Initial model loading depends on the network and device. Test one representative image before a large batch.

Related concepts

alpha channel
A per-pixel opacity channel where zero is transparent and the maximum is fully opaque.
segmentation mask
A model estimate of which pixels belong to foreground or background, used to compose transparent output.

Frequently asked questions

Does background removal upload my files?
No. Images are processed locally in your browser and are not uploaded to a server. The browser may download open-source model assets the first time you use it.
What format is the result?
The output is a PNG file with an alpha channel, useful for further editing, design work, or conversion to formats such as WebP.
Why is the first run slower?
The first run needs to load WASM and ONNX model assets. They are usually cached by the browser, so later runs in the same browser are faster.
Can it remove every complex background perfectly?
The tool automatically detects the foreground and creates a transparent background, but hair, translucent materials, low-contrast edges, or similar foreground and background colors may still need manual retouching.
Why are hair strands and fur left ragged?
Segmentation models decide foreground and background by region, and hair is a fine semi-transparent structure that is genuinely hard to resolve per pixel. Dark hair against a dark background is the worst case; reshooting against a contrasting background helps enormously.
Why is the first run so slow?
The first use downloads tens of megabytes of model files into the browser. They are cached afterwards, so subsequent runs in the same browser are much faster. Clearing site data invalidates the cache and forces a fresh download.

Related tools