Smart Crop

Centre crop, focal point and subject detection

Method	How the crop is placed	Fails when
Centre crop	Around the middle of the image	The subject is off-centre, as in most well-composed photos
Anchored crop	Against a chosen edge or corner: top, bottom left	Images in a set have different compositions
Focal point	Centred on one stored point, set by a person or by software	The new format is too narrow to hold the whole subject
Face detection	Around the detected faces	There are no people, or the faces are not the subject
Subject or saliency detection	Around the area a model judges most important	The model picks the wrong object, such as a bright background sign

A focal point (also called a focus point) is the simplest and most reusable of these. It is stored once with the image, often as percentages from the top-left corner, and every crop made later can use it. Because it is a percentage, it stays valid for every rendition size of the image.

How smart cropping works

1. Detection. A vision model or a saliency algorithm analyses the image and returns the position of the subject, often with a confidence score. Some systems return several points: the main subject and secondary ones, such as a product and the person holding it.
2. Fitting. For each target format, the crop box is made as large as the image allows at the required aspect ratio, then moved so the focal point is as close to its centre as possible without leaving the image.
3. Resizing. The crop is scaled to the exact pixel size of the format.
4. Review. A person checks the crops that matter and moves any that cut off something important.

The fitting step explains a common surprise: when the subject is at the very edge of a photo, the crop cannot be centred on it, because the box stops at the image border.

Where smart crop is used

• Social formats. One campaign photo becomes a 1:1 post, a 4:5 feed image, a 9:16 story and a 16:9 thumbnail. See creative automation.
• Responsive websites. The same hero image is shown wide on desktop and tall on mobile, with the subject visible in both.
• Templates. A product shot dropped into a banner layout fills its image frame around the product, not around the empty studio background.
• Grids and covers. Square tiles of portrait photos keep faces in view instead of showing shoulders.

When to crop by hand

Automatic crops are good defaults, not art direction. Crop by hand when:

• the image contains text, a logo or a legal line that must stay fully visible;
• the picture has two subjects that do not fit together in the narrow format;
• the crop changes the meaning, for example by cutting a person out of a group photo;
• the result will be printed or used in a paid campaign, where every frame is reviewed anyway.

Common mistakes include cropping from an already cropped export instead of the original, and enlarging a small crop to fill a large format, which shows as blur. For the files that crops produce, see rendition.

Smart crop in Freedam

• After upload, Freedam detects a primary focus point on every image, and on PDFs, vector files and office documents rendered as images. Secondary points are stored with their own positions and confidence, and a short description of each point is added to the search index.
• Social Export on an asset produces crops for preset formats such as Instagram Square Post, Instagram Story, LinkedIn Cover and YouTube Thumbnail. You choose which focus point the crops start from, or the image centre, then drag each frame to adjust it before generating. Each export is saved as a rendering on the asset. See renderings.
• In Studio, the Crop tool has Center on Focus, which keeps the crop box on the focus point, and Show Focus Area, which marks the point and its confidence. Turn Center on Focus off to place the crop freely. Dragging the crop past the image edge fills the new area with AI outpainting instead of cutting the subject.
• Image frames in templates can use the asset's focus point, prioritise detected faces, or fall back to an anchor such as top or centre.
• Collection covers, custom page covers and dashboard images are positioned on the focus point.
• Focus points are returned with each asset by the API.

Frequently asked questions

What is a focal point in an image?

A focal point is the position of the image's main subject, stored as coordinates. Cropping and resizing tools use it to keep the subject in the frame when the image is shown in a different shape.

Is smart crop the same as AI cropping?

Smart crop often uses an AI model to find the subject, but the crop itself is a simple calculation around that position. Generative tools that extend the image beyond its edges, called outpainting, are a different technique.

Can smart crop handle every image?

No. Images with several subjects, important text near the edges or a subject the model does not recognise still need a person to check or adjust the crop.