Best Photo Restoration for Genealogy and Family History Research

Most restoration marketing speaks to someone who wants a nicer picture above the fireplace. Family historians are after something stricter. You need a file you can put in a tree, a published book or an archive submission and still stand behind in ten years, long after you have forgotten the details of how it was made.

That reframes what counts as good work. A smooth, confident, perfectly sharp face is worse than a slightly rough one if the confident version is not actually your great-grandmother. A jaw eaten away by silver mirroring can be handled two ways. It can be read back out of the surviving emulsion, the light falling elsewhere in the frame and any other picture of the same person you can supply. Or the gap can be filled with something face-shaped and plausible. At thumbnail size both look fine. Only one is evidence.

Three questions separate them:

• What happens where the information is genuinely gone? An honest answer names the limit out loud. Automated tools close a gap by generating detail, and the file they return carries no marking to show which pixels survived and which were produced.
• Can somebody explain a decision? "I read the shadow under the collar and matched the hairline to the sibling portrait you sent" is a note you can file with the image. Software output has no reasoning to record.
• Does the invoice move with the damage? Research photographs are usually the worst photographs in the family. If a service prices by severity, your most important image is also your most expensive one.

MemoryCherish artists work in the United States, by hand, roughly four hours into a typical photograph, and the starting figure of $38 is decided by the tier you book rather than by the state of the print. For anyone working through a box of water-stained studio portraits, that predictability matters more than a slightly lower headline price.

Genealogy has a citation culture that photo restoration mostly lacks. Sources get recorded, evidence gets weighed, conclusions carry a confidence level. A restored image deserves the same treatment, and three habits are enough to give it one.

Keep the untouched scan permanently. The raw capture is your negative. File it beside the finished version with names that make the relationship obvious, something on the order of 1912-brennan-wedding_RAW and 1912-brennan-wedding_RESTORED, and back both up in two separate places.

Record the method next to the result. Who or what did the work, the date it was done, and what the restorer said could not be recovered. Two sentences in a text file will outlast your memory of the project by decades, and they are the difference between a documented image and a mystery.

Label anything the process generated. One company is unusually direct about this on its own help pages. MyHeritage states that because both the Photo Enhancer and In Color use deep learning technologies, the result is a "reimagination of what the original scene may have looked like". That is a vendor describing its own output, and it is the cleanest published description available of what an automated result actually represents.

Scope is worth checking too. MyHeritage's help center describes the Enhancer as upscaling photos to increase their resolution and automatically detecting faces in a photo and bringing them into focus. Higher resolution and a sharper face describe a different job from repairing a tear or a crease, so match any tool's published scope against what your photograph actually needs.

Hand restoration removes the labelling problem at source. MemoryCherish artists recover what is present in the photograph and never invent a face, so there is nothing generated to disclose. The original stays with you as well, which counts for a great deal when the print belongs to a cousin, a parish or a historical society that agreed to let you scan it once.

The person doing the work never met your ancestor. You have a census entry, a sibling's portrait, a regimental group photograph and a family memory of a scar above the left eyebrow. That knowledge shapes the finished file more than any price tier does, and most people never send it.

Put it in the order:

• Rank the frame. "The two faces matter, the studio curtain does not" prevents hours going into the wrong third of the image.
• Say what is not damage. Birthmarks, scars, a wandering eye, a chipped tooth, a missing fingertip. Retouchers are trained to clear blemishes away. Tell them which marks are the person.
• Attach references. A clearer picture of the same face ten years later is the strongest evidence anyone can hand a restorer about bone structure.
• State the destination. A half-page in a family history book and a framed print over the stairs need different resolutions and different decisions.
• Ask what could not be recovered. The honest answer is the useful one, and it belongs in your notes.

Then use the revision stage deliberately, and read each company's own wording rather than assuming a category norm. At the time of writing, Forever Studios states unlimited revisions are included with a restoration, and Rememorie states customers can request an unlimited number of free revisions after receiving the restored file. RetouchUp advertises free reworks. FixThePhoto's published workflow includes a step where the customer accepts the retouched photos or requests changes. Load each page the day you order, because the wording on the page is the policy.

MemoryCherish keeps revising at no charge until the result is right, which for research work usually means two or three passes at one stubborn face rather than starting the whole image again.

None of this argues for never touching an automated tool. Family historians deal in volume, and volume is precisely what software handles well. The discipline is keeping automated output out of the evidence chain.

Sensible uses:

• Triage. Push 200 shoebox scans through a fast pass to work out which six carry a face worth commissioning properly.
• Legibility rather than archiving. Reading a name inked on the back of a cabinet card, or a shop sign behind a street scene.
• Screen copies for a private tree, labelled as enhanced, with the raw scan stored alongside them.

What the vendors publish, all of it worth re-checking on the day you use it. At the time of writing, Hotpot offers its AI Picture Restorer free with no account needed, and states the free service repairs both color photos and black-and-white ones, automatically performing scratch removal, face enhancement and color sharpening. Hotpot also states on that same page that while the service automates photo restoration, it "cannot replace experts for demanding restoration jobs". Coming from the tool's own publisher, that is a useful description of where it sits.

VanceAI states its restorer model was trained on millions of vintage photographs to remove scratches, sepia tones, tears, cracks and spots, and that its restoration runs in one pass with no manual brushwork required. At the time of writing it states that sign-up credits cover three free restorations before a user picks a plan, that the free tier requires no credit card and applies no watermark, and that the web app restores one photo at a time on the free plan. Its pricing is credit-based, and its pricing page lists the AI Photo Restorer at 4 credits per image up to 4K and 8 credits at 4K.

Sort with these. Commission a person for the photograph that goes in the book.

Black-and-white film records luminance. It does not record hue. Any color applied afterwards is a decision taken by a person or a model, and no process recovers a color the emulsion never captured. That is physics rather than opinion, and it settles the question of how a colorized image should be catalogued.

It is not an argument against colorizing. It is an argument for labelling. MyHeritage says this about its own product, disclosing that photos colorized with In Color have "colors simulated by automatic algorithms, which may differ from the real-life colors". Treat that as the honest baseline for automated color: a plausible proposal, not a record.

Hand colorization is interpretation too. The difference is traceability. A person can research a choice (a regiment's facing color, a school's colors, the paint on the farmhouse that appears in a later photograph), can tell you which areas were researched and which were judgement, and can change them when you send a correction with a source attached. That gives you something to put in a footnote. An automatic output gives you nothing to cite.

Companies take positions here, and it is fair to report them as positions rather than findings. Forever Studios' colorization page contrasts its service with AI-based online colorization companies; that is their framing on their own page. At the time of writing, Forever Studios' site lists colorization at $55 per photo, and states the first set of colorization proofs arrives in about three days.

Whichever route you take, keep the black-and-white master as your file of record and treat the colored version as a display copy. If it goes into a book or a public tree, say that it was colorized and say by what method. A future researcher will thank you, and there is a decent chance that researcher is you.

The photograph that matters most to a genealogist is almost always the one in the worst condition. It is the only image of the immigrant grandmother, it spent thirty years in a wallet, and half her face has gone. Everything else in the box is easy by comparison. That one picture is where the decision is actually made.

We would hand it to a person. MemoryCherish is our own service, so weigh the recommendation accordingly, but the reasoning holds whoever you hire: several hours of human attention, artists working in the United States, a price that starts at $38 and follows the tier you select rather than the severity of the damage, free revisions until the face is right, a guarantee that returns cash instead of store credit, files that remain yours indefinitely, and a rule that only what survives in the photograph gets recovered. More than 400,000 families have ordered on those terms. Your original never leaves the house.

Reasonable alternatives, depending on the job in front of you:

• A digitizing backlog first. If the real obstacle is 900 slides and a reel of 8mm, a service built around bulk media transfer with restoration available on top will serve you better than sending images one at a time.
• A lower listed price per image. At the time of writing, Forever Studios' site lists $35 per image and RetouchUp's services page lists $35.00 per image, both below where we start. We would rather say that plainly than hope you do not check.
• Free triage passes before you commission anything, using the tools described above.

One caution worth stating flatly: some restoration sites publish no pricing, process or method information at all. For research work, where you may need to describe years later how an image was produced, a vendor you cannot document is a poor foundation. Choose someone who tells you what they do.