Remove the Background From an Image (Free, In Your Browser)
Removing a background from a photo is the most frequently-needed image edit nobody used to be able to do without Photoshop. You wanted a product photo cleanly cut out, a headshot dropped onto a different colour, a logo extracted from a screenshot — and the only options were "spend an hour with the lasso tool" or "pay for Photoshop."
That changed three years ago. Background removal is now a single neural-network pass that runs in milliseconds. The model is small enough to ship as a JavaScript download and run in any modern browser, on any modern laptop.
How it works under the hood
Modern background removal uses a segmentation model — a neural network trained to predict, for every pixel of an image, whether that pixel is "subject" or "background." The output is a mask: a black-and-white image the same size as the input, white where the subject is, black everywhere else. Compositing the original image with the mask as alpha gives you a transparent-background result.
The dominant model class is U²-Net (and its successors like RMBG, BiRefNet). Trained on hundreds of thousands of subject-on-background images with hand-labelled masks. The result is shockingly good for product photos, portraits, and anything with a clear figure-ground separation. It struggles with hair-on-busy-background, glass, transparent objects, and photos where the "subject" is ambiguous.
The free-tool tradeoffs
Search for "remove background free" and you'll get:
- Remove.bg — high quality, but free tier is 1 image/day at low resolution, then pay $0.20-1.99/image.
- Adobe Express — requires Adobe account, watermarks the free output.
- Canva — Pro feature, paywalled.
- Photoshop's "Remove Background" button — works, but you need a $20/month Adobe subscription.
- A dozen sketchy ad-laden sites — varying quality, all server-side uploads.
The reason free tiers are restrictive is server cost. Each background-removal run uses GPU time on the provider's infrastructure. Free tiers exist to drive paid conversions; the math doesn't work otherwise.
Browser-only background removal
The @imgly/background-removal library bundles the segmentation model and runs it via ONNX Runtime in the browser. First load downloads ~40 MB (cached after that); each removal then takes 1-3 seconds on a modern laptop, longer on phones.
Buncha's background remover wraps that library with the file-drop UI you'd expect. Drop an image, it processes locally, you download the transparent-PNG result. The image never uploads anywhere — verifiable by pulling up the network tab.
Quality is comparable to Remove.bg's paid tier. Resolution is limited only by your device's memory; we cap at 4K to avoid mobile devices stalling.
When to use it (and when not to)
The browser path wins when:
- You care about privacy. Product photos and headshots are exactly the kind of image you'd rather not have sitting in someone else's S3 bucket.
- You need batch processing. Cloud tools meter per-image; the local one runs as many as you want.
- You don't want an account.
It loses when:
- You need extremely-fine hair detail on a complex background. The cloud-side pro tools (Remove.bg, Photoshop) have larger models that handle the edge cases marginally better.
- You're on a phone and the model download (~40 MB) is unwelcome.
After background removal
A transparent PNG is the canonical output. From there:
- Composite onto a new background — drop the cutout into Figma, Canva, or any image editor on top of a coloured layer.
- Resize for a use case — Buncha's image resizer handles standard social media presets (Instagram, LinkedIn, etc.).
- Compress — transparent PNGs can be heavy. The image compressor handles them, though for the smallest result you'd convert to WebP.
- Add a watermark — for product photos especially, the image watermark tool stamps a logo over the cutout.
The whole sequence — original photo to background-removed-and-watermarked product image — runs entirely in your browser, no upload at any step.