The useful result
Create a cleaned-up copy of an old photo that keeps every person recognisable, with the original preserved.
- Create a cleaned-up copy of an old photo that keeps every person recognisable, with the original preserved.
A torn wedding portrait or a faded school photo is exactly what people hope AI can rescue. The risk is that a restoration tool treats missing detail as a guess and fills it with a plausible stranger’s face. Great-grandmother ends up with a different smile and nobody notices until a relative does.
This lesson keeps the repair honest: scan carefully, request cleanup only, and compare the result against the original in the places where invention is most likely.
The examples below are illustrative exercises, not recorded results from any tool. Sources were checked on 24 October 2026; menus, limits and prices change, so confirm details on the linked pages.
Before you start
- An old photograph or a good photo of it
- A phone or scanner
- An AI image editor that accepts photos
- A relative or friend who knows the people in the picture (optional but valuable)
Review evidence
Documentation-based review · 2026-10-24
Primary documentation and official guidance checked on 24 October 2026. Examples, checks and exercises are original illustrations, not recorded tool tests. Cover artwork is a conceptual illustration.
Scan once, keep the original, and check privacy settings
Scan or photograph the print flat, in even light, at the highest resolution available. Save the file as “original” and never edit it. Before uploading a photo of other people, read your tool’s data controls so you know whether uploads can be used to improve models and how to delete them.
Check your result
An untouched original is saved, and you know the tool’s upload and deletion settings.
Ask for cleanup only, and forbid changes to people
Request removal of scratches, dust and fading, and a gentle colour balance. State clearly that faces, expressions, hair, clothing, jewellery and background objects must stay exactly as they are. Avoid tools or settings that promise to “enhance faces”; these invent the most.
Restore this old photograph: remove scratches, dust and stains and gently correct the fading. Do not change any person’s face, expression, hair, clothing, jewellery or pose, and do not add or remove objects. If an area is too damaged to repair, leave it soft instead of inventing detail.Example / expected result
Illustrative example; not a recorded model result.
After cleanup, the creases are gone and the faces look like the original, only sharper in tone, not in features.
Check your result
Your instruction names what to clean and lists what must not change.
Compare at 200 percent and ask someone who knew them
Place the original and the result side by side. Check eyes, teeth, ears, jewellery, text on signs and the edges of clothing, because these are where details get invented. If anything looks new, undo that edit. Ask a relative whether the faces still look like the people. Label the file “restored (AI-assisted)”.
Check your result
Every face matches the original in the places listed, and the final file name says it was AI-assisted.
When it goes wrong
Colourising adds guessed colours that are not history. Upscalers can sharpen a face into a different face. Missing parts of a photo cannot be recovered, only imagined, so mark those areas clearly if you share the image. Keep the original forever.
Make it your own
Restore one low-stakes photo, such as a faded landscape or a family pet. Write two sentences on what the tool changed that you did not ask for, and show the before and after to someone.
Useful next resources
Sources and review
Last checked Oct 24, 2026. Features and account limits can change. Verify important results against your original material.
Did this help you complete your task?
Tell us what needs improving. No account or private output required.
Report incorrect information or a broken linkGenius Forges Editorial
Practical AI lessons with checked sources, labelled examples and human verification.
