How to Create a Portrait of Loved Ones Who Never Shared a Photo

Ask anyone who restores old photos what people really bring them, and repair requests are only half the story. The other half is a quieter wish: a grandmother who passed away before her grandchild was born, siblings separated by migration, a father who was always behind the camera and never in the frame. The family has photos of each person — just never together. The question is always some version of: can you put them in one picture?

For most of photography's history the answer was "yes, if you hire a skilled retoucher and wait a week." Today it's a few minutes on a phone. I build Heirloom, an AI photo restoration app whose Reunite feature does exactly this, so this guide covers both how the technology works and — just as important — how to use it honestly.

What a "reunite" portrait actually is

A reunite portrait is a composite: people from two or more separate photographs, combined into a single natural-looking image. Done well, it handles the hard parts that made manual composites look like collages — matching light direction, skin tones, film grain, scale, and perspective so the people look like they occupied the same room, not like stickers on a background.

It is worth being precise about what it is not. It isn't a recovered photograph, and it isn't a record of a moment that happened. It's a made object — closer to a commissioned painting of two relatives than to a snapshot. Families have commissioned exactly those paintings for centuries; the technology changed, the sentiment didn't.

Why this got dramatically easier

Three things converged. Modern restoration models can rebuild damaged or faded source photos first, so even a cracked 1950s print becomes usable input. Face-aware enhancement can sharpen a soft face without inventing new features — critical, because a memorial portrait where the person doesn't quite look like themselves is worse than no portrait at all. And compositing models learned to harmonise lighting and tone across sources automatically, which was always the giveaway in hand-made composites.

That last point is the quality bar to judge any tool by: if the two people appear lit from different directions, or one face is noticeably sharper than the other, the eye reads it as fake in milliseconds.

How to do it in Heirloom

The flow in Heirloom is deliberately simple, because the people doing this are usually families, not photo editors:

1. Pick a clear photo of each person. One good face per source photo is the only hard requirement.
2. Restore first if needed. If a source print is damaged or faded, run restoration on it before compositing — clean inputs make honest outputs. Your first restoration is free, so you can judge the quality before paying anything.
3. Create the Reunite portrait. Heirloom composites the people into one frame and harmonises light, colour, and grain. Faces are treated conservatively: the goal is that relatives say "that's exactly her," not "who is that?"
4. Export at full resolution. The output is print-ready — these images tend to end up framed, in memorial programmes, or shared across a family group chat within the hour.

On privacy, which matters more than usual here: photos are processed only to produce your result, never sold, and never used to train AI models. Grief is not training data.

Getting a result worth framing

From watching thousands of restorations, the difference between a moving result and a disappointing one is almost always the source photos:

• Face size beats photo size. A small print where the face is large outperforms a big group photo where the face is forty pixels wide.
• Match the era or embrace the mix. Two black-and-white sources composite beautifully; so do two colour photos. Mixing works best when you colourise the older photo first so the final image feels unified.
• Similar head angles help. Two faces both looking roughly at the camera will always composite more naturally than a profile and a portrait.
• Don't over-restore. Keep some grain. A slightly soft, filmic portrait reads as genuine; a plastic-smooth one reads as artificial.

The ethics: keepsake, not deepfake

This technology sits close to people's hearts, so the rules matter. The ones I hold Heirloom to, and would suggest to anyone making these images:

• Be truthful about what it is. Share it as "a portrait we made of Grandma and Maya," never as a found photograph. The emotional value doesn't depend on pretending — families cry over these knowing exactly what they are.
• Preserve likeness, don't invent it. Repairing damage and matching light is restoration; generating a smile that never existed on a face is fabrication. Tools should stay on the first side of that line.
• Consider the living. When the portrait includes living relatives, ask them first. Not everyone wants to appear in a composite, and the ask itself is a mark of respect.

Used this way, a reunite portrait is in the same tradition as the memorial paintings and retouched family portraits people have always made — a way of saying these people belong together, with better tools.

If you'd rather have a human do it

AI compositing isn't the only path, and it's not always the best one. For complex scenes — several people, specific poses, a particular background — a human retoucher on a commission platform can art-direct in ways an app can't, typically for $30–150 and a few days' turnaround. For the common case — two or three people, a clean portrait, this week's memorial — an app gets you a result in minutes for a fraction of that. There's no wrong choice; there's only what the moment calls for.

If you have a photo of each person and the wish to see them together, that's all you need to start — Heirloom is on the App Store, and the first restoration is free.