MemoryCherish vs Remini: What Happens to the Face

Nobody orders a restoration because they want cleaner pixels. They want to look at their grandmother and see their grandmother. That is the test this comparison lives or dies on, and it is why price and speed turn out to be secondary questions.

Old prints lose information. Silver tarnishes, emulsion cracks, a corner disappears behind a thumbtack, somebody folds a print to fit a wallet and the crease lands straight across a cheekbone. Once information is physically gone, something has to go in the hole. Every restoration method is a different answer to that one question: what do you put there?

Automated tools answer it statistically. Software fills a damaged region with detail consistent with the way faces of that type usually look, which is generation rather than recovery. Plenty of the time that is completely fine. On a mildly soft snapshot with lots of surviving detail, the output reads correctly and nobody would argue with it.

Hand restoration answers it forensically. A MemoryCherish artist rebuilds a damaged cheek from the intact side of the same cheek, from the grain and tone immediately around it, and from a second photograph of the same person if you have one to send. When the information genuinely is not recoverable, we say so instead of filling the gap with something convincing. We recover what the photograph actually holds, and we do not invent faces.

The people who catch the difference are rarely photographers. They are the ones who knew the person. They cannot always name what changed, only that something in the mouth or the set of the eyes is off, and that hesitation is the reason careful restoration exists at all.

On a holiday snap from 2011, the distinction is academic. On the only surviving photograph of your great-grandfather, it is the entire decision.

Their words, not our characterization, because ours would not be worth much here. Remini's photo restorer page describes its restoration as performed by AI technology. It describes the process as a one-click operation that removes blur, sharpens details and cleans up grain. Remini states its AI identifies faces and facial features in images, including blurry group photos from decades ago. Their homepage describes the Old Photos Restorer as bringing blurred, faded and damaged photos back to life, and at the time of writing lists it as a named tool among their enhancement tools.

Notice what that description optimizes for: faces, blur, grain, speed. Remini positions the product for family heritage photos among several use cases, and it also offers an API for integrating its enhancement technology into other products, which tells you something useful about the shape of the thing. It is built as an engine that runs at volume, and volume is a legitimate thing to build for.

Two limits on what we will say about it. We report what Remini publishes about its own product and stop there, because a company's published description is the only claim we can stand behind. And we are not putting a price on Remini in this article. Check the current terms in the app or on their site before you commit, particularly if you plan to run a batch of photos through it.

One phrase worth keeping straight: "one click" describes the interface and the compute behind it. It is not a delivery estimate for a human service, and reading it as one is how people end up comparing two numbers that were never measuring the same stage of anything.

Four hours sounds like a lot until you see the list. Here is roughly where it goes on an average photo.

• Assessment first. What is damage and what is the photograph? A white speck can be dust, a scratch, or a highlight on a brooch. Guess wrong and you erase the brooch.
• Structural repair. Tears, creases and missing corners get rebuilt from surrounding material, with the artist deciding what belonged there based on what the print still shows and what physics allows.
• Faces last, and slowest. Eyes, mouth corners and jaw lines carry the likeness. This is the part that eats the hours and the part a customer judges in the first two seconds.
• Grain and tone matching. Repaired areas get pulled back into the original texture, so the fix does not sit on the print like a smooth plastic patch.
• Your notes. If you tell us his left ear stuck out, it stays sticking out. That is not a small thing to the people who remember him.

Then the file arrives and the loop stays open. Revisions are free until the picture is right, rather than one free round followed by an invoice. If we cannot get there, the money comes back as cash instead of a store credit. The files are yours permanently. Your original never went anywhere, because we work from a scan or a good phone photo of the print.

The pricing works the same way on purpose. It starts at $38, and the artist tier you pick decides the figure, so a badly torn photo and a lightly faded one cost the same at the same tier. Nobody reviews your upload and revises the quote upward. Behind all of it: US-based artists rather than AI, more than 400,000 families served, and a verified review average of 4.96 out of 5 across 2,176 reviews and counting.

Remini makes sense when the print is structurally intact but soft, when you have a phone full of images from the digital-camera years, when you want to see something improved right now, or when the stakes are low enough that a slightly different-looking face costs you nothing. Their published description of a one-click operation that removes blur, sharpens details and cleans up grain is a fair match for exactly that kind of file.

MemoryCherish makes sense when the print itself is damaged, when the photograph is the only one that exists, when it is headed for a frame, a funeral program, a family book or a museum submission, or when a living relative is going to look at the result and rule on whether it is really them. Hand work is slower and it costs more. It also comes with a person who is accountable for the output and a revision loop that does not close until you say it does.

One comparison to refuse: an app's processing speed and an artist's working hours are not rival delivery estimates. The first describes how long software runs. The second describes how long a person spends making decisions about your specific photograph. Lining them up as though the smaller number wins is how people end up disappointed by a result they thought they had understood in advance.

If you are genuinely torn, the tiebreaker is usually replaceability. Can you rescan it, redownload it, or ask a cousin for a better copy? Then automate it. If the answer is no, put a person on it, and make sure that person will take your corrections afterwards.

Here is advice that costs us orders. Before you spend money anywhere, run your photo through something free and look at the output properly. Not a glance on a phone screen. Zoom to the eyes.

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. Hotpot states the tool automatically performs scratch removal, face enhancement and color sharpening.

Hotpot is also unusually direct about the boundaries of its own product, which is worth more than most marketing copy. Their page states that while the service automates photo restoration, it "cannot replace experts for demanding restoration jobs", and describes the tool as designed to help consumers with lightweight requirements while helping professionals save time on difficult restoration requests. A vendor telling you where its own line sits is genuinely useful information.

Two things to read before you use free output for anything permanent. At the time of writing, Hotpot states free images are limited by the CC BY-NC licence, and that it may use invisible watermarks to identify Hotpot creations for legal disputes. Hotpot states it retains images only as long as necessary to run its machine learning models. Its pricing page states prices vary by product and settings and are shown inside the product, with volume pricing of $0.02 to $0.25 per image for commercial-use credits, depending on volume, subscription and co-marketing.

Then judge the result. If it nails the photo, you are finished and you owe nobody anything. If the face came back smooth and slightly generic, you have learned exactly what your photograph needs, and you will brief an artist far better for having seen it go wrong.

Every claim above about another company is a statement published on that company's own site, checked at the time of writing in August 2026. Companies edit their pages. Treat these as dated snapshots rather than permanent facts, and click through if anything matters to your decision.

• Remini restoration method, one-click description and face identification: https://remini.ai/photo-restorer
• Old Photos Restorer product description, tool listing, heritage positioning and company identification: https://remini.ai/
• Hotpot free access, automated functions, licence terms, watermarking language and the expert-limits statement: https://hotpot.ai/restore-picture
• Hotpot pricing model and where prices are displayed: https://hotpot.ai/pricing

Remini's site identifies its operating company as AI Creativity S.r.l., Via Nino Bonnet 10, 20154 Milan, Italy, which is the kind of detail worth having if you ever need to contact a business about files of your own.

What we left out matters too. We have not stated a Remini price, and we have not described anyone's revision or refund terms unless the company publishes them. We describe no company's method beyond what that company says about itself, because what a business publishes is something we can point to and anything past that is guesswork dressed up as reporting.

If you are comparing services yourself, a habit worth borrowing: save a screenshot of the page you bought from, with the date visible. Published terms move, and a dated screenshot settles most disagreements before they start.