How to remove the background from an image

Get a clean transparent PNG for a product listing, a logo or a slide — without uploading your photo to anyone.

Background removal is the one image job that almost always ends with an upload. Most free tools require you to hand over the photo, and many then watermark the result or hold the full-resolution version back for a paid plan. Toolsfully does the work in your browser instead, so the picture never leaves your machine and the result comes back clean.

Two ways to do it, and when each one wins

The tool offers two methods because neither is best for every photo.

AI cutout uses a segmentation model that understands what a subject is. It traces around hair, plant leaves, chair legs and other awkward edges, and it copes with cluttered rooms where the background is a jumble of colours. The model is about 45 MB and downloads once, then stays cached — after that it works offline. Expect a few seconds per photo on a modern computer.

Plain colour needs no download and finishes instantly. It works inward from the edges of the picture, clearing everything that matches the border colour. For a product shot on a white sweep, a passport-style photo against a wall, or green-screen footage, it is both faster and cleaner than the AI. Its weakness is obvious once you know it: anything in the subject that shares the background colour goes too, so a white mug on white paper is a poor candidate.

Step-by-step

1. Open the tool. Go to Image Tools and pick the "Remove Background" tab.
2. Add your photos. Drag one in, or several — both methods run in batches.
3. Choose a method. Start with AI cutout unless your background is genuinely plain.
4. Pick what the background becomes. Transparent PNG is the default; you can also drop the subject straight onto white, black, a light blue, or a colour you choose.
5. Remove and download. A preview appears on a checkerboard so you can see exactly where the transparency falls.

Getting a better edge

• Shoot with separation in mind. A subject that contrasts with what's behind it cuts out well by any method. A dark jacket against a dark sofa is hard for software and for people.
• Use the tolerance slider deliberately. In plain-colour mode, raise tolerance when lighting across the backdrop is uneven; lower it when the subject contains colours close to the background.
• Add a little edge softness. One or two pixels of feathering stops the cut-out looking pasted on when you place it over a new background.
• Keep PNG for transparency. JPEG cannot store transparency at all — save as PNG unless you are flattening onto a solid colour anyway.

Frequently asked questions

Is my photo uploaded anywhere?

No. Both methods run inside your browser. The AI model is downloaded to your device and the inference happens locally, which is why the first run needs a moment and later ones don't.

Will there be a watermark, or a resolution limit?

Neither. You get the full resolution you put in, unmarked.

Why does the AI sometimes clip fine detail?

Segmentation models predict a boundary; wispy hair and semi-transparent edges are genuinely ambiguous. For a critical cut-out — a magazine cover, say — software of any kind is a starting point rather than a finish.