Free Image Cropper Online

Crop images with preset aspect ratios or a custom selection area. Download the cropped result instantly.

Your files never leave your device
Drop an image here or click to browse

Supports JPEG, PNG, WebP, GIF, BMP, TIFF, AVIF · up to 50 MB

How to Crop an Image

  1. Upload or drop an image above.
  2. Select an aspect ratio preset or use "Free" for any shape.
  3. Drag the crop selection to the area you want to keep. Use the corner handles to resize.
  4. Click "Crop Image" and download the result.

A Short History of Cropping

Cropping is older than digital editing by more than a century. To crop a photograph in the analogue era was a physical act: it meant deciding where to place a mask, an easel blade or a piece of card before the enlarger lamp exposed the print. In a working darkroom, an enlarger projects a negative through a lens onto a sheet of light-sensitive paper laid flat in an easel, a heavy metal frame with adjustable bands that overlap the paper's margins. Pulling the bands inward exposed only a smaller window of paper. For photo editors at newspapers and magazines, the workflow was even more material: editors marked their cropping decisions directly on a contact sheet using grease pencils (normally red but sometimes blue) drawing rectangles around the portion of each frame they wanted enlarged. The grease pencil was preferred because it wrote on glossy print emulsion without smudging. Henri Cartier-Bresson, the canonical Magnum figure, famously refused the practice, he insisted on full-frame printing and on composing inside the viewfinder at the moment of release, treating any later crop as intellectual cheating against the shutter. (His usually-cited exception is the 1932 image Behind Gare Saint-Lazare, where he had been shooting through a hole in a fence and physically could not move closer to fill the frame.) Ansel Adams stood at the other end: his meticulous darkroom practice treated cropping as one tool among many for shaping the final print, alongside burning, dodging, contrast filtering and selective bleaching. The toolbar icon for crop in modern image editors (two right angles bracketing an empty rectangle) is a stylised drawing of the L-shaped paper guides photographers used on contact sheets to frame potential crops by hand.

Aspect Ratios for the Platforms You Actually Post To

Aspect ratio describes the proportion of width to height. Modern social platforms each have one or more "preferred" ratios, and uploading at the wrong ratio almost always means the platform will crop or letterbox on your behalf, usually badly. A crop tool that gets you to the right ratio before upload is doing meaningful work. Confirm against current platform docs for production work (specifications drift) but as of 2026:

Cinema Ratios, Where the Wide Presets Come From

Academy ratio (1.375:1, often called 4:3 colloquially). Silent films were shot at roughly 4:3 using the entire 35 mm frame; when optical sound-on-film arrived in the late 1920s, the soundtrack stripe ate into the picture area. The Academy of Motion Picture Arts and Sciences responded with a directive on 16 February 1932 specifying new aperture dimensions, the result was a 1.375:1 frame, christened the Academy ratio, and almost every Hollywood film between 1932 and 1952 was shot in this format. CinemaScope (2.55:1 → 2.35:1 → 2.39:1). The technique behind it dates to the 1920s, when French inventor Henri Chrétien patented the Hypergonar anamorphic lens. 20th Century Fox signed an agreement with him on 12 January 1953 and announced on 2 February 1953 that all future Fox productions would use the technology; The Robe premiered in September 1953 as the first commercial CinemaScope feature. The original ratio was 2.55:1; when soundtracks were incorporated in 1956, it shrank to 2.35:1; SMPTE 195-1993 confirmed 2.39:1 as the modern standard. American widescreen flat (1.85:1): Universal Pictures introduced it on 20 May 1953 with Thunder Bay; non-anamorphic, achieved by hard-matting top and bottom. 16:9 (1.7777…:1) was first proposed in 1984 by Kerns H. Powers of the SMPTE HDTV Working Group, who chose it as the geometric midpoint between the cinema standards of the time. The ITU adopted it as the international HDTV standard in 1990; the FCC adopted ATSC in 1996; the first ATSC broadcasts went live on 28 October 1998. Every modern television, monitor, laptop screen, YouTube thumbnail and HDMI signal is still nominally 16:9.

The Rule of Thirds, A 1797 Composition Rule Still in Use

The phrase "rule of thirds" appears in John Thomas Smith's Remarks on Rural Scenery, published in 1797. Smith was an English engraver and antiquary; his book was ostensibly a collection of twenty etchings of cottages with text on the picturesque. In passing, Smith quotes a 1783 lecture by Sir Joshua Reynolds on the balance of light and dark in painting, then proposes to extend the same logic "to the rule of two thirds", coining what would centuries later become the standard composition heuristic for amateur and professional photography alike. Crop tools express the rule by overlaying a 3×3 grid on the live crop selection; aligning a horizon, a strong vertical (a tree trunk, a person), or a focal point with one of the four intersection nodes is the standard advice. Beyond thirds, several other composition aids have become standard overlays in serious crop tools: the golden ratio (≈1.618:1), sometimes described as the deeper mathematical inspiration of which the rule of thirds is a simplification, places its dividing lines at roughly 0.382 and 0.618 of the way across the frame; the Fibonacci spiral emerges from drawing successively smaller squares whose sides follow 1, 1, 2, 3, 5, 8, 13… and is what Renaissance painting compositions are often analysed against; the diagonal method / golden triangle draws a diagonal across the frame and a perpendicular from each opposing corner, suited to scenes with a strong oblique line. This tool does not currently render those overlays, a documented scope limitation and an honest opportunity for future work.

How the Crop Actually Happens, The Canvas API

The HTML Canvas API has, since the original WHATWG canvas spec, supported a single method that does the geometric heavy lifting of cropping: CanvasRenderingContext2D.drawImage(). It is overloaded with three forms, the third, nine-argument form is the one that crops: drawImage(image, sx, sy, sWidth, sHeight, dx, dy, dWidth, dHeight). The first four numeric arguments specify the source rectangle: a sub-region of the source image to extract. The next four specify the destination rectangle on the canvas where that sub-region should be drawn, with implicit scaling if the destination is sized differently. To crop without scaling, you size the destination canvas to match the source rectangle exactly and draw with dx = 0, dy = 0, dWidth = sWidth, dHeight = sHeight: which is precisely what this tool does. This single call is the beating heart of any browser-based crop tool. Once the pixels are on the canvas, the standard export path is canvas.toBlob(), which serialises the canvas to a Blob in a chosen MIME type (PNG, JPEG, WebP). The Canvas API also exposes a powerful adjacent function, createImageBitmap(), whose imageOrientation: 'from-image' option applies any EXIF-orientation rotation present in a JPEG before the bitmap is handed back, without it, photos taken on a phone held sideways would appear rotated 90° on the canvas. The 9-argument drawImage() has been baseline widely available since July 2015, meaning every browser in modern use supports it.

Lossless vs Re-Encoded: A JPEG Caveat

A small but important caveat for any browser-based crop tool: the Canvas API always rasterises and re-encodes. If you upload a JPEG, drop it onto a canvas, crop with drawImage() and export with toBlob('image/jpeg', 0.92), you have just performed one extra cycle of lossy DCT quantisation on top of the original. This phenomenon is known as generation loss, and JPEG is famously susceptible to it, each re-save introduces fresh quantisation error in the discrete cosine transform coefficients; chroma subsampling compounds the effect across saves, with chroma channels becoming progressively softer and colour boundaries bleeding. Cloudinary's well-known visualisation shows a JPEG re-saved fifty times degrading into a recognisable but visibly damaged caricature of itself. The standard advice is to keep an unmodified master copy of any image you might want to re-edit. There is, however, a special trick that allows a JPEG to be cropped without re-encoding, provided certain geometric constraints are met. JPEG compression operates on 8×8 blocks of pixels (or 16×16 macroblocks with chroma subsampling). If the crop boundaries fall exactly on these block edges (what the JPEG specification calls iMCU boundaries) the cropped region can be extracted by rearranging the existing compressed DCT coefficients without ever decoding them back to pixels. The canonical implementation is jpegtran, the command-line utility that ships with the Independent JPEG Group's libjpeg, organised by Tom Lane and first released on 7 October 1991. The Canvas API has no equivalent, a browser-based JPEG crop always re-encodes; the best a tool can do is choose a high quality setting (this tool uses 92%, which is conservative and visually competent). For PNG and WebP-lossless, the situation is different: those formats encode pixels deterministically, so a crop followed by a re-save in the same format is bit-perfect on the cropped pixels. PNG is the recommended output format for any crop where pixel preservation matters more than file size, for example, a screenshot or a UI mockup.

Touch and Pointer Events, Why the Handles Have to Be Big Enough

Until the Pointer Events specification matured, building a single drag-handle that worked correctly on a desktop mouse, a finger on a touchscreen and a stylus on a tablet required juggling three event families. Pointer Events Level 2 reached W3C Recommendation status on 4 April 2019, unifying these into a single coherent event family with pointerdown, pointermove, pointerup and a pointerType attribute exposing the underlying device class. For the user-facing parts of a crop tool (the corner handles you drag to resize the crop selection, and the centre region you drag to move it) touch ergonomics matter. The platform guidelines converge on a similar number: Apple Human Interface Guidelines specify a minimum hit-target of 44×44 points; Material Design specifies 48×48 dp; WCAG 2.5.5 (Level AAA, "Target Size") requires 44×44 CSS pixels; WCAG 2.5.8 (Level AA, "Target Size (Minimum)", added in WCAG 2.2) requires 24×24 CSS pixels minimum, with spacing exceptions. The practical implication: a corner handle painted as a tiny 8×8 pixel square is very hard to grab on a phone. The visible decoration can stay small, but the underlying hit area should extend out by transparent padding so the effective target meets at least the WCAG AA threshold of 24×24, ideally the 44×44 of the AAA / Apple guidance.

The Library Landscape, and Why This Tool Doesn't Use Any of Them

The dominant open-source client-side cropping library is Cropper.js by Chen Fengyuan, MIT-licensed, with roughly 13.8k GitHub stars and around 1.4 million weekly npm downloads as of early 2026. Its feature set is comprehensive: touch and mouse, mouse-wheel and pinch zoom, free-rotate, flip horizontal and vertical, multiple aspect-ratio modes, a movable image inside a fixed crop frame and rich event hooks. react-image-crop by Dominic Tobias is a pared-down React-specific component, ISC-licensed, under 5 KB gzipped. Pintura by PQINA is a commercial JavaScript image editor that includes cropping alongside resize, rotate, filter, annotation, sticker, watermark and redaction tooling, sold on subscription tiers from approximately $13 to $317 per month, used by companies including Dropbox, Square and ConvertKit. smartcrop.js by Jonas Wagner (2014) is a content-aware cropping algorithm that uses image-processing heuristics (colour saturation, edge detection) to find an aesthetically reasonable crop automatically rather than waiting for user input. This tool deliberately uses none of those, it is a small, single-purpose, framework-free implementation. It does not try to compete with Cropper.js on feature breadth; it is tuned for a single visit, a single image, a single crop, a single download, with no upload, no account and no script-tag pulled from a CDN.

Privacy: Why Browser-Only Matters Here

Server-side croppers (iLoveIMG, Adobe Express, Canva and most "online image cropper" sites) require uploading your image to a third-party server. iLoveIMG documents a two-hour deletion policy; Adobe Express and Canva have similar but variable terms. Even with the most scrupulous deletion, the upload itself is a network event that can be logged, intercepted or breached. For ordinary holiday photos this is fine. For personal photos with EXIF GPS metadata, document scans containing PII (passport pages, ID cards, tax documents, insurance forms), screenshots of internal work UI, medical imaging or anything else you would not want copied onto a stranger's hard drive, "the image stays on my device" is the only safe architecture. This tool runs the entire crop pipeline (file selection, decode, canvas crop, re-encode, download) locally in your browser via JavaScript. No upload, no API call, no log entry. You can verify by opening DevTools' Network tab while you crop: there are no outbound requests carrying image data. Better yet, take the page offline (airplane mode) after it loads and confirm the cropper still works, the strongest empirical proof that nothing is being uploaded.

Frequently Asked Questions

Does cropping reduce image quality?

For PNG and WebP-lossless: no, cropping is bit-perfect on the cropped pixels. For JPEG: yes, slightly. The Canvas API always rasterises and re-encodes, so a JPEG-in-JPEG-out crop adds another generation of lossy quantisation. This tool uses a 92% quality setting on JPEG output, which is conservative, most viewers will not notice the difference, but if you plan to re-edit the crop later, keep an unmodified master copy. For pixel-perfect crops on screenshots, UI mockups or any image where every pixel matters, choose a tool that can output PNG or WebP-lossless.

Can I crop to exact pixel dimensions?

The crop selection shows its pixel dimensions in real time, so you can drag until the readout matches your target. For exact pixel sizes (1080×1080 for an Instagram square, 1200×630 for a Facebook OG image), use the Image Resizer tool after cropping to scale the result to precise dimensions. The combination (crop here, resize there) covers every social-platform export workflow.

Why is my iPhone photo appearing rotated 90°?

iPhones (and most modern phones) capture in the sensor's native orientation and write an EXIF orientation tag (1-8) telling viewers how to display the image. Older versions of the Canvas API ignored that tag, so phone-shot photos sometimes appeared sideways in browser-based croppers. Modern browsers (Chrome 81+ from 2020 onward, Firefox and Safari following) apply EXIF orientation by default to <img> elements, and createImageBitmap() does the same with imageOrientation: 'from-image'. If you still see a rotated photo, your browser is unusually old, update it.

Which aspect ratio should I pick?

Match the destination platform. Instagram feed engagement favours portrait 4:5 (5:4 vertical, 1080×1350) over square 1:1; YouTube thumbnails are 16:9 (1280×720); TikTok and Instagram Stories/Reels are 9:16 vertical (1080×1920); Facebook/LinkedIn link previews via Open Graph are 1.91:1 (1200×630); Pinterest standard pins are 2:3 (1000×1500). For print, A-series paper is 1:√2 (≈1:1.414); 35mm film and most DSLR sensors are 3:2; iPad and older 4:3 photos are 4:3. Use "Free" if you are cropping for art rather than for a specific platform.

Is my image uploaded anywhere?

No. All processing (file selection, decode, crop selection, canvas re-encode, download) happens locally in your browser via JavaScript and the HTML Canvas API. No image is sent to a server, no API call is made, no log entry is created. You can verify in DevTools' Network tab while you crop, or take the page offline after it loads and confirm the cropper still works. Safe for personal photos with embedded GPS, document scans containing PII, screenshots of internal work UI or any image you would not want copied onto a stranger's hard drive.

Does cropping strip EXIF metadata?

Yes, by side-effect. The Canvas API only stores pixel data, so EXIF, IPTC and XMP metadata blocks (camera model, exposure settings, GPS coordinates, copyright tags) do not survive the round-trip, they are gone from the cropped output. This is a privacy win when you are sharing the cropped image to forums or chat apps that don't strip EXIF on upload. If you need to preserve metadata (photographers archiving exposure data, content workflows requiring copyright tags), this is the wrong tool, use ImageMagick's convert or a dedicated EXIF-aware library that explicitly preserves metadata.

Related Tools

Image Compressor

Compress images up to 80% smaller, right in your browser.

Image Resizer

Resize images to exact pixel dimensions.

Image Converter

Convert between PNG, JPEG, and WebP formats.