Combine Family Photos

Start with two photos containing clearly visible faces, preferably taken from a similar camera height. Similar lighting direction and a comparable view of the upper body make a believable composition easier. A tiny face cropped from a distant group photograph provides much less useful information than a sharp original portrait. Avoid reference images with heavy filters that change facial proportions.

Use one person per reference where possible. If a photo contains more than one person, spell out which people belong in the final image and count them before generating. This preset begins with two reference photos. Guest generations support up to five sources. Begin with a small group and check each face: more photos do not guarantee accurately rendered people.

The button sets image-to-image mode, multi-image fusion and a 4:3 output ratio. Add at least two distinct reference photos. The workspace now blocks a multi-image submission with fewer than two sources and explains what is missing. A normal single-image edit remains available if you decide that a reference already contains everyone you need.

Opening the preset only fills the working settings; it does not upload photos or start a generation. After the upload completes, verify that every intended reference appears in the source list. Review the prompt and the generation cost shown for the selected model. Do not submit again just because a processing task has not produced its result immediately; use the current task state first.

The default prompt uses reference order from left to right. Make that instruction more specific when needed: the adult from the first image on the left and the adult from the second image on the right, both at similar scale. Refer to visible clothing or upload order instead of names the model cannot infer from the photos. State the final person count in your edited prompt.

Keep the first composition simple: a neutral background, comfortable spacing and soft shared light. Adding complex embraces, overlapping hands or a busy dining table increases the number of anatomical and spatial decisions the model must invent. If the photos have very different body crops, request a head-and-shoulders composition that fits both references rather than asking it to reconstruct unseen full bodies.

Count the final people before assessing the overall style. Compare each face to its own reference, watching for blended features, duplicated faces and a missing member. Check the hairline, glasses, expression and visible clothing. A correct group count does not prove that the model assigned the right reference to each position.

Next inspect hands, shoulders, body scale and the space between people. Shared lighting and contact shadows should make sense across the group. If one person looks much larger, request consistent camera distance and a simpler arrangement. If faces blend, shorten the prompt and distinguish references by order and visible features. Return to the original references for another attempt rather than repeatedly editing a composite that already lost an identity.

Download an approved image and keep the individual references separately. Let recipients know that the scene is a creative composite when there could be confusion about the event. This workflow does not provide evidence that people met, attended a ceremony or stood together in a real location. It also does not validate personal identities.

For a frame or a card, check the crop before printing. The 4:3 setting provides a wider starting canvas; it does not certify a physical print size. The browser image cropper can prepare a destination-specific version after you approve the composition. Do not crop away a person's hair or shoulders merely to fit a template without first checking the result at the intended display size.