Removing the background from a photo used to mean twenty minutes with the pen tool in a desktop editor. Today an AI model can separate a subject from its surroundings in a couple of seconds. This guide walks through the whole process — from choosing a good photo to exporting a clean file — and explains what to do when the automatic result needs a touch-up.
What "removing the background" actually does
Behind the scenes, a segmentation model looks at every pixel and estimates how likely it is to belong to the main subject. The output is a mask: a grayscale image where white means "keep", black means "remove", and shades of gray mean "partly transparent". Those in-between values matter — they're what make hair, fur and soft edges look natural instead of cut out with scissors.
The mask is then applied to your original photo as an alpha channel. Where the mask is black, the pixels become fully transparent; where it's white, they stay exactly as they were. Nothing about the subject itself is redrawn, which is why the colors and detail of the person or product are untouched.
Step 1: Start with a photo that helps the AI
Any photo can be processed, but a few habits make the result noticeably cleaner:
- Contrast between subject and background. A dark jacket against a dark wall is hard for anyone — human or machine — to separate. If you're taking the photo yourself, choose a background that differs in color or brightness.
- Sharp focus on the subject. Motion blur and soft focus smear the edge, so the AI has to guess where it is.
- A clear main subject. One person, one product or one car works best. Crowded scenes with several overlapping objects may need manual cleanup.
- Enough resolution. More pixels along the edge means a more precise mask. Use the original file rather than a screenshot or a messaging-app copy when you can.
Step 2: Remove the background
Open the background remover and add your image in whichever way is easiest:
- Drag the file onto the page.
- Press Ctrl + V (or Cmd + V on a Mac) to paste an image you've copied.
- Click Browse to pick a file — JPG, PNG, WebP, HEIC from an iPhone, and AVIF all work.
- Paste a direct link to an image (this works when the website hosting it allows it).
The first time you use the tool, your browser downloads the AI model. That can take a little while on a slow connection, and the progress bar shows exactly how far along it is. After that the model is cached, so the next images start almost immediately — even without an internet connection.
Step 3: Check the edges
Use the compare view to check the result. Drag the slider across the image and look closely at the places that usually cause trouble:
- Hair and fur
- Thin parts such as straps, cables, bicycle spokes or plant stems
- Holes inside the subject — the space between an arm and a body, the inside of a handle
- Areas where the subject and background had a similar color
Zoom in with the mouse wheel or a pinch gesture. Most photos need nothing at all, but it's worth a ten-second look before you export.
Step 4: Refine what the AI missed
Open the Refine panel to correct anything that isn't right:
- Restore brush paints back parts that were removed by mistake.
- Erase brush removes leftovers such as a patch of floor or a slice of wall.
- Size, hardness and opacity control the brush. Use high hardness for crisp, geometric edges like a phone or a logo, and low hardness with lower opacity for hair and fabric.
- Feather softens the entire outline by a few pixels, which hides tiny jagged steps.
- Edge shift grows or shrinks the cut-out. A value of −1 or −2 is the quickest way to remove a thin halo of the old background color.
Every change can be undone with Ctrl + Z, and Reset returns you to the original AI result.
Step 5: Choose what goes behind the subject
A transparent background is only one option. In the Background panel you can switch to:
- A solid color — white for marketplaces and ID photos, or a brand color for social posts.
- A gradient — subtle gradients look more premium than flat color in thumbnails and banners.
- Your own image — place a person in front of a different location or a product in a styled scene.
- A blurred version of the original — the fastest way to a "portrait mode" look while keeping the real setting.
If the subject looks pasted on, add a soft drop shadow in the Effects panel. Pointing it straight down with a large blur and low opacity makes the subject feel grounded.
Step 6: Export in the right format
- PNG keeps transparency and is lossless. Use it for logos, design files and anything you'll edit further.
- WebP also keeps transparency but produces much smaller files — ideal for websites.
- JPG has no transparency, so it's for images with a solid or photo background. It's the most compatible format for uploads.
We cover the trade-offs in detail in PNG vs JPG vs WebP for cut-outs.
Why in-browser processing matters
Many online background removers upload your photo to a server, process it there, and send the result back. That's fine for a holiday snapshot, but less comfortable for client work, product prototypes, ID photos or pictures of your children.
This tool works differently: the model runs inside your browser using WebAssembly, or WebGPU when your device supports it. Your photo is never uploaded anywhere. You can verify this yourself by opening your browser's developer tools and watching the network tab while an image is processed — the only downloads are the model files, and nothing containing your image goes out.
Frequently asked questions
Is there a size limit? Files up to 25 MB are accepted. Editing uses a fast preview, but exports are rendered from your full-resolution original.
Can I process lots of photos at once? Yes — batch mode processes many images in a queue and downloads them all as a ZIP.
Do I need an account? No. There's no sign-up, no watermark and no limit on the number of images.
Ready to try it? Remove a background now — it takes about as long as reading this sentence.