Image Cropper — Free ratios, rounded and circular masks, batch cropping, all local
Cropping an image means drawing a box and throwing away everything outside it. Start by getting the box right: free crop lets you pull any width and height you like, or you can lock to the usual ratios — 1:1, 4:3, 16:9, 9:16, 21:9 — or type your own width-to-height pair and keep it as a template. Then shape it. A rounded rectangle takes a radius in pixels or percent, and each of the four corners can be set on its own; the circle mask gives you a true circle on a square selection and an ellipse on a rectangular one; the polygon mask offers a hexagon, an octagon and a heart. These shapes only change the outline you see and the outline that gets exported — the crop underneath is always rectangular. On export a second Canvas pass masks it: PNG gets a transparent outside, while JPG has no alpha channel, so the outside is filled with a base colour you pick. The image itself can be turned ±90°, straightened to any angle to rescue a tilted shot, and flipped horizontally or vertically. Output can follow the selection's own pixels or be forced to a width and height you name, with a no-upscale switch so a small original is never stretched into mush. When you are done you can slide a before/after divider, hold the magnifier over the result to check for jaggies on a rounded edge, and crop a whole batch at once into a ZIP, with a coordinate JSON for re-cropping on the server. Everything runs in the browser — nothing is uploaded, and it works on phone, tablet and desktop.
Import
Crop canvas
Cropped result
Crop ratio
FreeWith a locked ratio the handles resize proportionally only. Switch back to Free to pull it any way you like
Shape mask
Transparent only means anything for PNG / WebP / AVIF. Exporting to JPG fills white, or whatever base colour you pick
Output size
If the original does not have the pixels, do not force it — that only ends in mush
Output formatApplies to JPEG / WebP / AVIF only. Higher is sharper and heavier
SmoothingExport re-encodes the pixels, so EXIF such as camera model, GPS and capture time does not travel with the file
File name
Use {w} and {h} for the output width and height
How to use
- Drop the image into the box above, or click Select image to pick several at once, or just press Ctrl+V to paste a screenshot. To pull an image from a URL, paste it into the link field. Phone photos are rotated upright from their EXIF tag automatically, so there is no need to turn them first.
- Drag out a crop box on the canvas — drag inside it to move, drag any of the eight handles to resize. To lock a ratio, click a preset on the right such as 1:1 or 16:9, or type your own width and height; switch back to Free to pull it any way you like. The wheel zooms the whole image, dragging empty space pans it, and a two-finger pinch does the same on a phone.
- Add a mask if you need one: a rounded rectangle with an adjustable radius (corners can be set separately), a circle that becomes a true circle on a square selection, or a polygon — hexagon, octagon or heart. A mask only changes the outline. On export PNG stays transparent outside, while JPG is filled with the base colour you chose.
- Finally set the output size, format and quality, then use the panel on the right to slide a comparison or hold the magnifier over the result. Download when happy, bundle a batch into a ZIP, and export a crop-coordinate JSON plus a CSV manifest if you need a paper trail.
Frequently asked questions
I applied a rounded or circular mask, so why is the exported image still the full rectangular size?
Because a mask never changes the selection — it only turns the pixels outside it transparent. The crop area is still that rectangle, so the exported width and height are the rectangle's width and height. If you want dimensions that hug the shape, make the selection square first and then apply the circle mask; or keep the alpha channel by exporting PNG and let whatever consumes the file trim to alpha.
I picked the circle mask and the JPG came out with white corners. Did I set something wrong?
No — JPEG simply has no alpha channel. Every pixel outside the circle has to land on a real colour, which defaults to white. You can switch it to black or a custom value under Fill the empty area with. If you genuinely need transparency, export PNG, WebP or AVIF; all three carry alpha.
Photos from my phone arrive sideways. Do I need to rotate them first?
No. The EXIF orientation tag is read during decoding and applied immediately, so every rotate and straighten you do afterwards works on an already-upright image — there is no way to end up back on its side. That tag is discarded on export, so the file will not be rotated a second time by other software.
I turned on Never upscale, so why is the output smaller than the size I typed?
That switch means: when the original does not have the pixels, do not force it. If your selection is only 300×200 and you ask for 1200×800, it exports at 300×200 instead of blowing the image up four times into mush. If you really do want the upscale, just turn the switch off.
In a batch, can I adjust just one image?
Yes. Ratio, mask, corner radius, output format and quality are shared across every image, but clicking any row switches to that image so you can drag its crop box on its own — the adjustment survives switching away and back. To re-fit every image to the shared ratio at once, click Reset to shared ratio.
What is it
Cropping looks like nothing more than cutting off the part you do not want, but it is really two operations stacked on top of each other: choosing a rectangular region, and deciding what shape that region should take. The first is ancient — in a darkroom you just slid the easel blades inward and blocked the light. The second had to wait until pixels could carry transparency. So every rounded corner, circle, hexagon and heart in this tool is not cut out; it is masked out on a transparent canvas. The rectangular region is resampled and painted to fill the frame first, then a path acts as the blade, keeping only the pixels inside it. That also explains a common misunderstanding: a mask never changes the crop underneath. The circle you see is a rectangle with its four corners covered, and the exported dimensions are still that rectangle's width and height. JPEG has no alpha channel, which is why choosing a circle and exporting to JPG requires filling the outside with a real colour — otherwise those pixels get crushed to black.
Origin & history
Before digital imaging, cropping was a manual job in the darkroom. Under the enlarger sat a metal easel that held the paper down; you pushed its four blades inward, and only what was left uncovered got exposed. That was the first crop box, and it came with a limitation that is hard to imagine now — you could not preview the result, only estimate it. From the 1950s, newspaper picture editors began marking crop lines on the back of a print, two short strokes per edge showing what to keep, and photographers trimmed to match. Those marks became the crop marks of the printing trade, and eventually the CropBox entry inside every PDF. When Photoshop 1.0 shipped in 1990, cropping was still a plain rectangle: drag a box, press Enter, the rest was gone, with no option beyond undo. What turned cropping into something you could keep revising was the arrival of layers and non-destructive editing — Photoshop 3.0 added layers in 1994, 4.0 added adjustment layers in 1995, and people began keeping the original and treating a crop as a decision they could revisit. Rounded corners travelled a different road. In 1995 CSS could only draw square boxes, so a rounded corner meant slicing nine images and stitching them back together; the CSS3 draft proposed border-radius in 2005, and browsers only agreed on it around 2011, after which the rounded corner became the cheapest and most overused decoration in interface design. What made arbitrary shapes routine was two things: clipPath in the 1999 SVG specification, and clip() added to canvas in the 2004 HTML5 draft. The latter let a Bézier path act as a blade inside the browser, cutting a rectangular canvas into a heart or a hexagon — and that is exactly the route every rounded export from this tool takes. One footnote: the fix for sideways phone photos was written into the EXIF standard back in 1998, as the Orientation field. Most software simply could not be bothered to read it, and only around 2019 did browsers start honouring it while decoding.
Related keywords
Feedback
Found a problem with a tool or have a suggestion? Email us — we read every message.
