Load your image
Drop in a JPG, PNG, or WebP file — or try the built-in sample. Nothing leaves your device.
Background removal guide
Cutouts used to mean paid software or uploading your photos to a stranger's server. FitMyPic does it locally: click the background to sample its color, fine-tune tolerance and edge feathering with live preview, then download a real transparent PNG — or swap the background for white, black, or any custom color in one click.
Know your image
Workflow
Drop in a JPG, PNG, or WebP file — or try the built-in sample. Nothing leaves your device.
The color under your cursor becomes the removal target. Adjust tolerance for near-matching shades and feather to soften edges, watching the checkerboard preview live.
Keep true transparency, or pick white, black, or a custom fill color. The measured output size is shown before you download the PNG.
Two removal modes
Removes only background regions touching the image border. Safer when the subject contains a similar color — a white shirt on a white backdrop keeps its shirt.
Removes all pixels matching the sampled color anywhere in the image. Use it for cutouts with holes (between an arm and torso) or repeated background color in separate areas.
Questions
Yes. The exported PNG contains a genuine alpha channel. The checkerboard in the preview is only a convention for showing transparency.
Photos with solid or near-solid backgrounds: white studio product shots, flat-color backdrops, and lightly uneven tones. Busy or highly detailed backgrounds need professional AI segmentation tools.
Tolerance decides how similar a pixel must be to the sampled color to become transparent. Feather softens the transition at edges so cutouts don't look harshly outlined.
Yes. Choose solid white, solid black, or a custom color in the Replace with menu — feathered edges blend smoothly into the new background.
No. Everything runs locally in your browser with the canvas API. Your images are never transmitted to a server.
Next: resize your cutout for Instagram or see how we test output.
Removing a background by colour matching is a bet: the algorithm assumes the subject is separable from what surrounds it because the two differ in colour. That bet pays off in a predictable set of cases and fails in an equally predictable set.
It works well on: images photographed against a plain, evenly lit surface (product shots on white paper, a person against a painted wall), logos and flat illustrations on a single colour, scans of documents on clean backing, and screenshots of graphics where the background is one fill.
It fails, or needs manual work, on: gradient backgrounds and studio lighting with shadows, hair and fur, semi-transparent objects (glass, smoke, veils, tulle), reflective or shiny edges picking up the background colour, busy surroundings, and anything where the subject shares colour with the background.
Half-transparent pixels are the boundary case worth understanding. An antialiased edge contains pixels that are genuinely mixed between subject and background. A hard threshold has to decide they are one or the other, and that decision is where the ugly border comes from.
The white fringe that appears around a cutout is almost always one of three things.
Trapped background pixels. The alpha mask stops one or two pixels short of the actual subject boundary, so a thin band of background colour survives at the edge. Against a different backdrop the band reads as a rim of the old background.
Colour bleed from a scaled image. If the source was resized or compressed with JPEG before the cutout, edge pixels carry colours blended toward the background. Those blends survive as a coloured halo even when the alpha is correct.
A matte colour that does not match the destination. When a transparent image is composited onto white, any subtle semi-transparent edge tint shows. The same file placed on a dark background may look perfectly clean.
Fixes, in order of effort: reduce tolerance so fewer background pixels are kept; add a slight feather so the boundary is a gradient rather than a hard line; and inspect the result against both a white and a dark checkerboard before shipping.
Check the cutout over two different backgrounds. An edge that looks perfect on white can show a dark halo on a dark surface, and vice versa. This is the failure mode that is easiest to miss and hardest to diagnose later.
The single largest quality gain usually comes before any tool touches the file: start from the uncompressed original. Edge pixels in a JPEG that has already been re-encoded at quality 60 have colour bleed built in that no tolerance setting can undo.
| Source | Expected quality of a colour-based cutout |
|---|---|
| Original PNG or high-quality export, plain background | Clean edges, minimal halo |
| JPEG q80+ on a plain background | Good at normal viewing sizes, some fringing at 100% |
| Heavily recompressed JPEG, or a screenshot of a photo | Visible fringing that survives every setting |
| Photo with shadow or gradient background | Requires manual masking, not colour matching |
Some jobs look like background removal but are not:
Zoom to 200% on the subject's edge and look for the old background's colour. If it is there, lower the tolerance and re-export. If it survives at the lowest setting that still removes the background, the image needs manual masking.
The background remover on this page reads pixels locally, applies tolerance and feathering settings you control, and produces a PNG without uploading anything — so you can test settings against the original file as many times as you like.
Source: background separation described in the PNG specification's transparency model — https://www.w3.org/TR/png-3/
Everything in this guide runs locally in your browser — free, no account, no uploads.
Open the free tools