Skip to content
QuickToolio
esEspañolenEnglishptPortuguêsfrFrançaisdeDeutschitItaliano

Automatically cut out a person from a photo

Cut a person out of a photo with AI and keep only them: transparent background and a canvas trimmed to their silhouette, with no empty margins around it. Everything happens in your browser.

Processed on your device via WebAssembly. Nothing is sent to any server.

or click to select it

A segmentation model (u2netp, the lightweight variant of U²-Net) works out which pixels belong to the person and applies that mask as transparency. The canvas is then reduced to the smallest rectangle containing the person: if they only fill the middle of a landscape photo, the final file is only that area. That is the difference from "Remove Background from Image", which returns the cutout at the photo's original size.

Use cases

  • Placing a person on a résumé, badge or team-page template without trimming margins by hand.
  • Making an avatar or sticker with a transparent background.
  • Adding someone to a slide or collage taking up only the space of their silhouette.

How it works

  1. 1

    Upload the photo

    Any image your browser can open (JPG, PNG, WEBP...), up to 25 MB and 40 megapixels.

  2. 2

    Click "Remove background"

    The first time, about 19 MB is downloaded (the AI model and the engine that runs it); while you stay on the page, the next photos are near-instant.

  3. 3

    Download PNG or WEBP

    Both formats keep the transparency and come already trimmed to the rectangle the person occupies.

Technical details

To trim the canvas, mask pixels with at least 6% opacity (16 out of 255) count as the person: this discards near-invisible background noise without cutting the semi-transparent edges of hair. No margin is added. If the model finds no subject, the image is returned at its original size.

The model is not person-specific: it detects the main subject. If the person is holding an object, is with a pet or several people stand together, they are usually all kept and the canvas fits the whole group.

Analysis runs at 320×320 pixels and the mask is scaled back to the original resolution, so the cropped area keeps its original pixels, but very fine edges can lose detail. EXIF orientation from phone photos is respected.

Frequently asked questions

How is this different from "Remove Background from Image"?

They use the same model. "Remove Background" keeps the original size and position, which is useful if you'll overlay the result on the original photo; this tool also removes the transparent margins and gives you only the person's rectangle.

Why does the result have more empty space than expected?

Because the model kept a fragment of the background with some opacity away from the person, and the canvas fits everything that was kept. You can fine-tune the framing afterwards with "Crop Image".

What if there are several people in the photo?

If they're close together the model treats them as one subject: they're cut out together, not separately.

Related tools

← Image