How to remove the background from an image
Remove the background from a photo in your browser, get a transparent PNG or a plain colour, and fix the edges when the cut-out isn't perfect.
· 3 min read
To remove the background from a photo, drop it into Remove Background: an AI model finds the subject and erases everything else, and you download a transparent PNG or the subject on a new plain colour. It works best on portraits; products and animals on a fairly plain background usually come out well too.
The quick way
- Open Remove Background and drop your photo.
- Under New background, choose Transparent, White or Colour (then pick the colour).
- Choose the Output format: PNG (transparency), WebP (transparency, smaller) or JPG (no transparency).
- Click Remove background, compare the original and the result, then download.
The first time, your browser downloads a small AI model (about 7 MB), so the first image takes a moment longer. The browser then keeps the model in cache. Your photo is processed on your device and is not uploaded.
Transparent or plain colour?
- Transparent is for placing the subject on something else: a slide, a website, a poster, a logo layout. Use PNG or WebP — JPG can't store transparency, so the tool switches to PNG if you pick JPG with a transparent background.
- White is the usual choice for product photos on marketplaces and for catalogue pages.
- Colour is handy for profile pictures or ID-style photos that need a uniform backdrop. Check the exact requirements of the organisation first: official photo rules often specify the background colour, the framing and sometimes forbid edited photos.
Get a cleaner cut-out
The result depends mostly on the photo you start with.
- Contrast helps. A dark subject on a light wall, or the reverse, gives cleaner edges than a beige jumper against a beige sofa.
- Keep the subject whole and in focus. Blurry edges and cropped limbs confuse the model.
- Crop first if the subject is small. If the person or object only fills a corner of the picture, crop the image around it before removing the background. The model then works on the part that matters.
- Use the best version you have. A heavily compressed, low-resolution image gives rough, blocky edges.
Common problems
Hair or fur looks cut out. Fine strands are the hardest part for any automatic tool. The result is usually fine at normal viewing size; for a large print you may need to touch up the edges in an image editor.
Glass, smoke or thin objects disappear. Transparent and very thin things (a glass of water, a bicycle wheel's spokes, a veil) are often partly removed. That is a limit of automatic background removal, not something a setting will fix.
Part of the background stayed. This happens when the background looks like part of the subject — a chair someone is sitting on, a shadow, a hand holding the product. Busy scenes with several people or objects are harder too. Try a photo where the subject stands apart more clearly.
The model won't load. The first use needs an internet connection to download the model. If it fails, check your connection and try again.
After removing the background
- Need another format? Image Converter turns the result into JPG, PNG or WebP. Keep in mind that JPG fills transparent areas with a solid colour.
- File too heavy for a form or email? A transparent PNG can be large. Compress Image reduces it; WebP output is usually lighter to begin with. For more on this, see how to reduce image file size and JPG vs PNG vs WebP.
Limits worth knowing
The tool removes the background and can replace it with a plain colour. It doesn't place the subject on another photo or add shadows, and it has no manual brush to repaint the edges. For those, open the transparent PNG in an image editor.
Tools used in this guide
- Remove Image BackgroundCut out people and objects with AI — transparent PNG.
- Crop ImageCut an image to the area and aspect ratio you need.
- Image Converter (to PNG, JPG, WebP, BMP, ICO)Convert any image to PNG, JPG, WebP, BMP or ICO.
- Compress ImageShrink JPG, PNG and WebP images without visible quality loss.