1. Home
  2. Images
  3. Background Remover

Background Remover

Remove photo backgrounds with an AI model that runs in your browser.

Runs in your browser · No upload

How to use

  1. Drop photos onto the box or click to choose them (up to 20 at a time). You can also paste a screenshot with Ctrl+V.
  2. Click Remove background. The first time only, a 4.6 MB AI model is downloaded and saved on this device. After that, nothing is downloaded. Each photo takes a few seconds, depending on your device.
  3. Adjust Threshold and Soften edges while you watch the result, and use the brush to bring back removed areas or erase leftover background.
  4. Choose a result background (transparent PNG, solid-color JPG, or a photo behind the subject as JPG) and download. With several photos, you can get them all in a ZIP.

Options

Option What it does Range
Threshold How strictly to cut the model’s “this pixel is part of the subject” score 0 to 100%, default 50%
Soften edges Blurs the alpha (transparency) edge with a Gaussian blur 0 to 4px (for a 1000px long side)
Brush, restore or erase Fixes the automatic result by hand. Undo with Ctrl+Z, redo with Ctrl+Shift+Z Size 1 to 30%, hardness 0 to 100%
Shadow Adds a soft black shadow under the subject Strength 5 to 100%
Trim empty margin Crops the transparent border, then adds back a margin set as a % of the subject’s long side 0 to 30%
Result background Transparent (PNG), a solid color (JPG), or your own photo behind it (JPG, scaled to fill)

How the threshold works

The model gives every pixel a value from 0 to 1. For a threshold T, pixels at or below T × 0.75 become fully transparent, pixels at or above that value plus 0.25 become fully opaque, and values in between blend smoothly. Here’s what model values of 0.2, 0.4, 0.5, 0.6 and 0.8 turn into:

Threshold 0.2 0.4 0.5 0.6 0.8
20% 0.104 1 1 1 1
50% 0 0.028 0.5 0.972 1
80% 0 0 0 0 0.896

If faint shadows or the tips of hair disappear, lower the threshold. If specks of background remain, raise it.

Large photos, margins and backgrounds

  • A photo with a side longer than 4,000px is scaled down, keeping its proportions, before processing. For example, 6000×4000 becomes 4000×2667 and 4032×3024 becomes 4000×3000.
  • Edge softening scales with the long side: a value of 1 is 3px on a 3000px photo and 0.5px on a 500px photo.
  • On a 2000px photo, the shadow is shifted 30px down with a 40px blur. The default strength is 40%.
  • Adding a 5% margin around a 400×600 subject adds 5% of the long side (600), which is 30px, on every side. The result is 460×660.
  • A background photo is cropped around its center so it fills the frame with no gaps. A 1600×900 photo placed in a square frame uses its central 900×900 area.

What it works well on

People, pets, products and food: photos where the subject is clear. The model is small and lightweight, so it has limits. Hair tips, transparent objects (like a glass), smoke or shadows, and areas close to the background color can come out rough or partly cut off. In those cases, change the threshold or touch up with the brush. It doesn’t promise the quality of professional commercial tools.

Privacy and storage

Your photo is processed only in your browser (in a web worker) and is never sent to a server. The model file is downloaded from this site and kept in browser storage (Cache Storage). Its SHA-256 is checked on download, so a damaged file is rejected. You can delete saved models at any time with “Delete saved models”. The model is U²-Net (the small u2netp version), licensed under Apache-2.0.

To erase just one object, use Remove Objects from Photo. To enlarge a photo, use the AI Image Upscaler. To make it smaller, use Resize Image.

Your photos are processed in your browser and never sent to a server.

Can't find the tool you need?

Tell us what you want to do. We build the most requested tools first.