Crop Image Online — Free Image Cropper

What cropping actually does to your pixels

Cropping is the one image edit that removes data instead of transforming it. When you crop a 4000 x 3000 photo (12 megapixels) down to a 1000 x 1000 square, you are not compressing or resampling — you are discarding the 11,000,000 pixels outside the selection and keeping the 1,000,000 inside it exactly as they were. That is why cropping never softens the part you keep: a 1:1 selection covering 1000 x 1000 of the original comes out as a true 1000 x 1000, pixel-for-pixel identical to that region of the source, each pixel still 24 bits across 3 channels of 8 bits. This tool cuts from the native resolution of the file you uploaded, not from the smaller preview shown on screen, so the export is as sharp as the original allowed.

Under the surface it is a single HTML5 canvas operation. The cropper reports the selection as 4 numbers — an x and y offset plus a width and height, all in source-image pixels — and the tool draws that rectangle onto a fresh canvas sized to match, using the 2D context's drawImage with its 9-argument form: source x, source y, source width, source height, then destination x 0, y 0, and the same width and height. Because destination and source dimensions are equal, there is no scaling step and no interpolation — the kept pixels are copied 1 to 1. The canvas is then exported with toBlob in your chosen format.

Aspect ratios and the platforms that demand them

An aspect ratio is the proportion of width to height, and most platforms reject or auto-mangle anything that does not match theirs. The 6 options here cover the common cases. 1:1, a perfect square at 1.0:1, is the Instagram profile picture (320 x 320 stored, 110 x 110 shown) and the classic 500 x 500 product thumbnail. 4:3, or 1.33:1, is the traditional photo and slide proportion inherited from 35mm and early 640 x 480 screens; 3:4, or 0.75:1, is its portrait flip for standing subjects. 16:9, the widescreen standard at roughly 1.78:1, is the native size for a YouTube thumbnail (1280 x 720), a 1080p frame (1920 x 1080), a 4K frame (3840 x 2160), and most presentation decks. 9:16, about 0.56:1, is that same ratio stood on end — 1080 x 1920 — the default for Instagram Reels, TikTok, and Stories. Free mode removes the constraint entirely so you can trim to any rectangle.

Locking a ratio changes how the crop box behaves: it keeps those exact proportions while you drag to reposition and use the 1x to 3x zoom to frame tighter. That guarantees the export lands in the target slot without the platform stretching it — an off-ratio image uploaded to a 16:9 slot is either letterboxed with bars or centre-cropped by an algorithm that does not know where your subject is. Cropping to spec first keeps that decision yours. A practical habit borrowed from photography is the rule of thirds: mentally divide the frame into a 3 x 3 grid and place the subject on a gridline or intersection rather than dead centre, which usually reads as more balanced.

PNG, JPG, or WebP: picking the right export

The format you export decides the trade-off between fidelity and file size. PNG is lossless: it keeps every one of the cropped pixels exactly and supports a full 8-bit alpha channel, 256 levels of transparency, so it is the correct choice for logos, screenshots, graphics with hard edges, or any image with a transparent background. Its cost is size — a photographic crop can run several times larger as PNG than as JPG. That is why PNG has no quality slider here: there is no quality to trade.

JPG is the opposite: built for photographs, it uses lossy compression to reach the smallest files, but it has no alpha channel at all. If your source is transparent and you export JPG, the tool fills the canvas white first so those areas do not turn black. WebP is the modern middle ground, delivering JPG-class or better compression while still supporting transparency. For JPG and WebP a quality slider runs from 50% to 100% and defaults to 92%, a level that is visually lossless for most photos while still shrinking the file; drop toward 50% or 60% only when size matters more than the last traces of detail. As a rough guide, a 1000 x 1000 crop can land near 300KB as JPG at 92, near 900KB as WebP with alpha, and past 2MB as PNG. The preview shows the exact output width and height in pixels before you commit.

Zoom, resolution, and a common misconception

The 1x to 3x zoom slider is a framing aid, not an enlarger. Zooming in makes the crop box cover a smaller region of the source, so the output has fewer pixels, not more — it never invents detail to fill a larger canvas. A 3x zoom on a 3000-pixel-wide photo frames roughly a 1000-pixel-wide slice (3000 divided by 3), and that slice exports at about 1000 pixels wide, sharp because those are real source pixels; a 2x zoom would frame about 1500, and a 1.5x zoom about 2000. This is the opposite of upscaling, which stretches few pixels across many and softens the result. If you need a small crop to end up physically larger, that is a job for an AI upscaler after cropping, not for the zoom control.

Because everything runs on the canvas in your browser, the whole cycle — decode, crop, encode — takes a fraction of a second even for a 24-megapixel file, and the image never touches a network. The practical limit is memory: an extremely large source is held uncompressed as RGBA at 4 bytes per pixel — a 24-megapixel photo occupies about 96MB in RAM before encoding — so a 100-megapixel panorama near 400MB can strain a phone or a 4GB laptop. For ordinary photos, screenshots, and web graphics that ceiling is never in sight, and you get an instant, private crop at true resolution with the format and size shown before download.

Quick reference

• Aspect presets: Free, 1:1, 4:3, 3:4, 16:9, 9:16 (6 options).
• Zoom: 1x to 3x, framing only — never upscales.
• Output formats: PNG (lossless, alpha), JPG (smallest, no alpha), WebP (compressed, alpha).
• Quality: 50–100% slider for JPG and WebP (default 92%); PNG is always lossless.
• Resolution: cut at the source's native pixels, not the on-screen preview.
• Engine: HTML5 canvas drawImage (9-argument form) + toBlob.
• Transparency: preserved by PNG and WebP; JPG fills white behind transparent pixels.
• Privacy: 100% in-browser; the image is never uploaded.
• Readout: exact output width x height shown above the preview.
• Cost: $0, unlimited, no watermark, no sign-up.