Smart Crop - Content-Aware Automatic Image Cropping

Pick a target aspect ratio and the tool works out which part of the photo to keep, scoring candidate crops by edge detail, colour saturation and skin tones so the subject stays in frame. Runs entirely in your browser.

Analysis runs on your device. The photo is never uploaded.

How content-aware cropping decides what to keep

A centre crop is the default almost everywhere, and it is wrong surprisingly often. Photographers habitually place subjects off-centre, so squaring up a landscape shot by chopping equal amounts from each side regularly slices someone in half or leaves a face at the very edge of the frame.

Content-aware cropping takes a different approach. It slides a window of your chosen aspect ratio across the image and scores every candidate position on three signals: edge density, which marks out detail and structure against empty sky or blurred backgrounds; colour saturation, since vivid regions usually carry the subject; and skin-tone detection, which weights faces and people heavily. The highest-scoring window wins, and there is a mild preference for placing that region near the classical rule-of-thirds points rather than dead centre.

The practical result is that a wide product photo cropped to a square keeps the product, a group shot cropped to a 9:16 story keeps the faces, and a landscape cropped to 16:9 keeps the horizon interest instead of a strip of empty sky. It is particularly handy when you are producing the same asset at several ratios, since you can re-crop from the original each time rather than nudging a box by hand.

All the scoring happens in JavaScript on your own machine, using the open-source smartcrop library. There is no model call to a server and no upload, so the tool is just as usable on client work, personal photos and internal material as on anything else. The crop is taken from the full-resolution original, so you get real pixels rather than an upscaled preview.

How to auto-crop a photo

  1. Upload the photo. Use the largest version you have, the crop is taken from the original pixels.
  2. Choose a target ratio. 1:1 for avatars and product grids, 16:9 for headers and video thumbnails, 9:16 for stories and Reels, 4:3 or 3:4 for print-style layouts.
  3. Click Auto Crop. The scoring pass runs in a fraction of a second and the chosen region appears on the canvas.
  4. Download. Save the cropped image, then repeat with a different ratio if you need the same shot in several formats.

Need to place the crop box yourself instead? The manual image cropper gives you a draggable selection. For sizing an image to a specific platform's dimensions after cropping, use the social media resizer.

Features

Subject-aware scoring

Edge density, colour saturation and skin-tone weighting decide the crop, so faces and products stay in frame.

Runs on your hardware

The analysis is plain JavaScript in your browser. No model server, no upload, no usage tracking.

Five common ratios

Square, 16:9, 9:16, 4:3 and 3:4, covering avatars, thumbnails, stories and print layouts.

Frequently asked questions

How does smart cropping decide where to crop?

It slides a window of your chosen ratio across the image and scores each position on edge density, colour saturation and skin tones, with a mild preference for rule-of-thirds placement. The highest-scoring window is the crop.

Does it use AI or a machine-learning model?

No, and it does not need to. It uses the open-source smartcrop library, which applies classical image-analysis heuristics. That is why it runs instantly on your own device with no model download and no server call.

Can I crop the same photo to several sizes?

Yes, and this is where it saves the most time. Change the ratio and click Auto Crop again, each crop is recalculated from the full-resolution original rather than from the previous result.

What if I want to choose the crop myself?

Use the manual image cropper, which gives you a draggable selection box and exact pixel control.

Are my photos uploaded?

No. The image is read locally and analysed on a canvas in your browser. No sign-up, no usage tracking, and no copy on a server.

Does the AI upload my image?

No — the model runs in your browser. The saliency detection is executed on your own device using the browser's canvas and WebAssembly/WebGL, so the photo is only ever read into local memory. Nothing is sent to an inference server, there is no queue and no copy of your image is retained anywhere; close the tab and it is gone.

🔎 Verify it yourself: nothing uploads

Don’t take our word for it. Open your browser’s DevTools (F12Network tab), then run this tool on any image. The only download you’ll see is the open-source AI model coming to your browser — your image itself is never sent anywhere. Every tool on this site works the same way — all processing happens on your device.