The 30-second method
Use the freeimgtool Background Remover. Drop the photo, pick Transparent as the output, click Remove background, download.
The tool runs an instance segmentation model in the browser. Nothing is uploaded to a remote server — the image stays on your device, the model processes it locally, and the result downloads as a clean PNG with an alpha channel.
- Open the Background Remover.
- Drop a portrait or product photo.
- Pick Transparent (default).
- Click Remove background.
- Click Download for a transparent PNG.
What photos work best
Portraits with a clear separation from the backdrop. Product shots on a plain background. Single subjects with high contrast against their surroundings.
Photos with busy or low-contrast backgrounds need cleanup in a desktop editor afterwards — the automatic cut won't be perfect.
Even lighting helps the model identify edges. A subject lit from one direction with shadow on the same color as the background is the hardest case — the model has no contrast to anchor the edge. If you control the shoot, place the subject at least a meter from the backdrop and use a soft fill light on the shadow side.
Solid color backgrounds (white, neutral gray, chroma green) produce the cleanest cutouts. Patterned backgrounds with similar tones to the subject's clothing or hair produce the messiest. If you have a choice, shoot against the simplest backdrop you can.
Fixing rough edges
If the edge looks ragged or the background bleeds in, paste the cutout onto a contrasting background to see exactly where the issue is, then touch up in any free pixel editor (Photopea is web-based).
The two common artifacts are edge halos and missed regions. Edge halos appear as a thin fringe of the original background color around the subject — that's because the model softened the alpha at the boundary. The fix is to nudge the alpha channel: increase contrast on the alpha mask, or contract the selection by 1-2 pixels.
Missed regions usually happen with hair, glass, or anything semi-transparent. The model is forced to make a hard call: subject or background. For wispy hair you can recover detail by running a second pass with a higher edge feather, or by manually painting back transparency on the affected region with a soft brush.
Real-world use cases and stats
Marketplace product photos. Sellers on Etsy, eBay, and Amazon report click-through rates 10-30% higher on listings with white-background cutouts versus busy-context photos. The cutout pulls the eye to the product without the visual clutter of a kitchen counter or coffee table.
Profile pictures and team pages. A consistent transparent or solid-color background across a team page makes the page look intentional. The alternative — each person photographed in their own room with their own lighting — looks chaotic even if every individual photo is fine.
Composite design work. Removing the background once gives you a reusable asset you can drop into mockups, social posts, ads, and slide decks for the next year. Each individual reuse takes seconds instead of minutes of manual masking.
File size impact: a 2 MB JPG portrait becomes a 600-900 KB PNG with transparency after cutout, because the new alpha-zero regions compress almost for free in PNG's deflate stream. That's a noticeable bandwidth win for product catalogs at scale.
When the automatic cut won't be enough
Hair against busy backgrounds. The model can usually find the main hair silhouette but loses individual strands. For magazine-quality cutouts where each strand matters, plan on manual cleanup in a layer-based editor.
Transparent objects: glass, water, smoke. There's no clean boundary to detect because the background shows through the subject. The model usually treats the entire transparent region as background. For these subjects, a chroma-keyed shoot or manual masking is the only reliable path.
Multiple overlapping subjects. The model decides which is the foreground based on its training prior. If you want two people in the cutout and one is partly behind the other, the result is usually correct, but if you want only one of them, expect manual work.
Reflective surfaces like jewelry and metal. The reflections often pick up the background color, which the model can interpret as either subject or background. Results vary; check the edge against a contrasting background before relying on the cutout.
Tools and alternatives
Remove.bg is the well-known commercial alternative. It produces excellent results but watermarks the high-res output on the free tier and requires payment for full-quality downloads. The freeimgtool version is free at full resolution.
Photoshop's Select Subject runs a similar segmentation model and integrates with manual refinement tools (Refine Edge, Select and Mask). For one-off professional work where edge quality matters, Photoshop's manual cleanup is hard to beat.
GIMP and Photopea (both free) have manual selection tools (foreground select, intelligent scissors) that take longer but give precise control over each edge. For complex cutouts with hair detail, a hybrid approach works well: use an automatic tool for the main silhouette, then clean up edges manually in GIMP or Photopea.
For batch processing, rembg is an open-source Python library that runs the same kind of segmentation model locally. Useful when you have hundreds of product photos to cut out and want a scriptable pipeline.
Frequently asked questions
Is the output watermarked?
No. The downloaded PNG is clean.
Does it work on people with long hair?
Edges around coarse hair are smoothed. Wispy hair on a busy background is the hardest case.
Can I get white background instead?
Yes. Pick White (or any solid color) instead of Transparent in the output picker.
Will it work on multiple subjects?
Yes. The model segments all foreground subjects together. If you want only one of several people, the automatic cut won't separate them — you'll need to crop or mask manually afterwards.
Is there a file size limit?
The browser-side model handles photos up to about 4000 px on the long edge comfortably. Beyond that, you may want to resize before running the cutout to keep processing fast.
Can I use the result commercially?
The tool itself adds no restrictions. The underlying photo's license terms still apply — make sure you have the rights to the original image.