LeanImg

Remove Background — Free AI Background Remover

Remove backgrounds from images instantly with AI. Get transparent PNGs for product photos, portraits, and graphics — free, no sign-up, no watermarks.

Loading background remover...

Want to thank us? Tell your friends about LeanImg!

How to Remove Backgrounds

1

Select your images

Choose files or drag and drop them into the background remover.

2

Remove backgrounds

Click the button and the AI will automatically detect and remove backgrounds.

3

Download

Download the transparent PNG images individually or as a ZIP.

Why LeanImg?

Removing backgrounds is essential for product photography, social media content, presentations, and graphic design. Most online tools either charge per image, add watermarks, or limit free tiers to low-resolution previews. LeanImg is different — our background remover is powered by IS-Net, a state-of-the-art neural network that delivers results on par with premium services — and it runs entirely in your browser, so your images never leave your device. It is completely free, requires no sign-up, and processes full-resolution images with no watermarks. Explore all of our free image tools to optimize your workflow.

State-of-the-Art AI

Powered by IS-Net, a leading neural network for image segmentation. Handles hair, fur, and semi-transparent edges with precision — 100% on your device.

Free, No Limits

No per-image fees, no watermarks, no sign-up. Full-resolution output, unlimited usage — completely free.

No Software to Install

Remove backgrounds online in seconds. No downloads, no sign-ups — just drag, drop, and go.

What runs in your tab

Background removal here is a neural network executing on your own hardware. The model is IS-Net, run through the IMG.LY background-removal library under AGPL-3.0, and the model files are served from this site rather than a third-party CDN. Inference uses WebGPU where the browser offers it and drops to CPU where it does not, which is slower but produces the same cutout.

The first run downloads a model

This is the part that hosted tools never have to mention, so here is the number: a cold run pulls about 190 MB across roughly 52 requests before it can start. That is the actual trade. A server-side remover uploads your photo, which is small, and sends a small result back, and the cost is that a stranger's machine has held your image. This one never sends the image anywhere and downloads the machinery instead. On a fast connection you will barely notice. On a metered one, that is a real number and you should know it before you begin.

When the cutout comes back empty

A saliency model needs something salient. Hand it a landscape, a texture or a flat-lay with no clear subject and it can return a PNG that is almost entirely transparent, which is easy to mistake for success because the file gets smaller rather than larger. So every result is measured before you see it: the tool samples the alpha channel and, if less than 0.1% of the frame survives at meaningful opacity, it tells you no subject was found instead of presenting an empty file as a finished cutout. For scale, a real portrait measures around 43% by the same statistic. The two outcomes are nowhere near each other.

Why the result is a bigger file

Output is always PNG, and that is forced rather than chosen. Transparency needs an alpha channel and JPEG has none in any mode, so a cutout saved as a JPEG would arrive on a solid rectangle. PNG is lossless, so a photograph that squeezed into a few hundred kilobytes as a JPEG can come back several times larger. The fix, if you need one, is the compressor's palette mode, which quantizes the colors and keeps the alpha channel intact.

Frequently Asked Questions

How does LeanImg remove backgrounds?

LeanImg uses IS-Net, a state-of-the-art neural network for image segmentation, running entirely in your browser. Your images never leave your device — the AI identifies the foreground subject and separates it from the background locally, producing a transparent PNG. The model handles complex edges like hair, fur, and semi-transparent objects with high accuracy.

What file formats does LeanImg support for background removal?

You can upload JPG, PNG, WebP, AVIF, GIF, BMP, and ICO images. The output is always a transparent PNG file, which preserves the cutout with full alpha transparency. You can then compress or convert the result using our other free image tools.

Can I remove backgrounds from multiple images at once?

Yes. LeanImg supports batch background removal — select up to 10 images at a time and process them all in one go. Download results individually or all at once as a ZIP file. No sign-up or account required to use batch processing.

Is the background removal really free?

Yes, completely free with no limits. There is no sign-up, no watermarks, and no hidden fees. Unlike competitors that charge per image or cap resolution on free tiers, LeanImg processes full-resolution images at no cost. Use it for product photos, portraits, or any image.

How accurate is the background removal?

LeanImg uses IS-Net, a neural network built for dichotomous image segmentation. It excels at fine detail like hair strands, transparent objects, and complex edges. Results are comparable to premium paid services like remove.bg and PhotoRoom — with the difference that everything runs privately on your own device.

Which model does the background remover use?

IS-Net, run through the IMG.LY background-removal library, which is licensed AGPL-3.0 and credited in the footer. The model files are served from this site rather than a third-party CDN, so using the tool does not hand your IP address to anyone else. Inference runs on WebGPU where the browser supports it and falls back to CPU where it does not.

Why is the first run slow?

Because the model has to arrive before it can run. A cold run pulls about 190 MB of model and runtime files across roughly 52 requests. That is the honest cost of doing this on your device instead of on a server: a hosted tool sends a small image up and gets a small image back, while this sends nothing and downloads the machinery instead. On a metered or slow connection that trade is worth knowing about before you start.

Why is the cutout a bigger file than the original?

Because the output has to be a PNG. Transparency needs an alpha channel, JPEG has no alpha channel in any mode, so a cutout cannot be a JPEG without losing the thing that makes it a cutout. PNG is lossless, so a photograph that compressed well as a JPEG becomes considerably larger as a PNG. If the result needs to be smaller, run it through the compressor in palette mode, which keeps the alpha channel.

What happens if my image has no clear subject?

You get a warning rather than a silent failure. A landscape or a texture with nothing salient in it can come back as an almost entirely transparent PNG, which is easy to mistake for a good cutout because the file even gets smaller. Every result is measured: if less than 0.1% of the frame survives at meaningful opacity, the tool says so. A healthy portrait measures around 43% by the same statistic, so the two cases are not close.

Is there a resolution limit?

There is no pixel cap. The tool has been verified on inputs up to 256 megapixels, and a 5760x5075 portrait comes back at full resolution with no downscaling. The limits that do exist are on the batch: up to 10 images per run and 50 MB in total. Very large images take longer and use more memory, since the work happens on your own hardware.

Which file formats can I use?

JPG, PNG and WebP go straight through. GIF, AVIF, BMP and ICO also work, decoded by the browser first. SVG and HEIC are rejected with a clear message rather than a cryptic library error, because Chrome cannot decode either one in this context. Whatever goes in, a transparent PNG comes out.

Does my image get uploaded?

No. The only network requests are the model and runtime files from this site, and they travel in one direction. Your image is read from disk into the page, handed to the model in memory and written back out as a blob you download. There is no upload endpoint in the codebase to receive it. Open your network tab during a run: you will see model chunks arriving and nothing carrying your photo away.

Have more questions? Learn more about LeanImg and how we process your images.

Need to compress your images? Compress your images to reduce file size. Or convert between formats like JPG, PNG, WebP, AVIF, and more. Want to resize? Resize your images to exact dimensions.