← All guides · 2026-10-05 · PhotoUnfade team

How Much Does Photo Restoration Cost? A 2026 Price Guide

Restoring a damaged family photograph in 2026 costs anywhere from nothing to $300 or more, and the deciding factor is almost never quality — it is who does the work and how much of the image is physically missing. An AI service costs cents per photo, a freelance retoucher typically charges $25 to $75, a specialist studio quotes $100 to $200 and up for severe damage, and a paper conservator stabilizing the actual object is a different profession at a different price. Across a 200-photo shoebox those routes differ by a factor of a thousand or more, which makes the budget question really a triage question: which handful of photos is worth human hours, and which 195 are not.

How much does photo restoration cost per photo in 2026?

Here are the five routes families actually use, with the typical published price for each. Read the last column first — matching the route to the damage saves more money than shopping any single route for a discount.

Two notes. A restoration that looks simple often is not: a single clean tear through a face costs more than a photo covered in light scratches, because the retoucher must build plausible features where none survive — which is why most studios quote only after seeing the image. And the DIY row is the one people misprice most often: the software is cheap, the hours are not.

Where do these numbers come from?

Every figure here is one of two things, and we have kept them separate. Measured numbers come from our own production system and our own invoice: for this article we queried our live job table, which holds 30 completed restorations with recorded processing times and output dimensions. Market ranges are published prices for professional restoration, digitizing and desktop software collected on 5 October 2026 — typical ranges, not a quote from any named vendor. If you spot a number you cannot trace to one of those two categories, email hello@photounfade.com and we will either source it or delete it.

What makes a restoration quote go up?

Six variables drive most of the spread between a $25 job and a $250 job. Describe all six when you request a quote and you will get comparable answers.

1. Missing content, especially over a face. The single biggest cost driver. Repairing what is still there is fast; rebuilding what is gone is slow and skilled.
2. Number of faces. Every face is checked individually — a wedding photo of eleven people is not one job, it is eleven.
3. Colorization. Hand-colorizing adds hours of judgment about era-appropriate fabric, skin and hair tones; automated colorization adds essentially none — which is why it should be a mode you choose knowingly, never a side effect of clicking restore.
4. Output size. Restoring for a wall print demands more than restoring for a 4×6. See below — this is where people get caught.
5. Rush turnaround. Commonly a 25 to 100 percent premium.
6. Input quality. A glare-covered phone photo costs more to work with and produces less. Poor input is the most common reason a project gets paid for twice.

What does it cost to restore a whole box of 200 photos?

This is the arithmetic that changes the decision, because almost nobody with a shoebox has just one damaged photo. At the published market rate of $25 to $75 per photo, hand-restoring all 200 costs $5,000 to $15,000. The hybrid approach — machine for the volume, human for the handful that genuinely need reconstruction — costs roughly a tenth of the low end.

The last row surprises people: for a large box, the scanning step can cost more than the restoration step. If you can borrow a flatbed scanner, digitizing yourself at 600 dpi is both cheaper and safer for fragile prints — our digitizing guide gives the exact settings, including the automatic enhancements you must switch off because they are destructive and cannot be undone later.

What does AI restoration actually cost to run?

Our own numbers, from our invoice and our job table, because the gap between compute cost and retail price is why this category is priced the way it is. One restoration costs us ¥0.18 — about $0.025. The content-safety classification pass that screens every upload adds roughly ¥0.002 and 0.2 to 0.5 seconds. Wall-clock time averaged 22.7 seconds across our 30 completed jobs, ranging from 5.8 to 33.2 seconds depending on image size. So at $9.9 for 100 credits we collect about $0.099 per photo against roughly $0.025 of compute; at $29 for 1,500 credits, about $0.019.

Which brings us to a disclosure we would rather make ourselves than have someone else find: if an annual subscriber used all 1,500 credits, our compute cost would be about ¥270 — roughly $38 — against $29 of revenue. We would lose about $9 on that customer. Like most plans of this shape, ours is priced on the assumption that most people restoring a family archive use 30 to 200 credits, not 1,500. We think that is fair and we monitor it, but you are entitled to know it exists, because a plan that loses money on heavy use eventually gets quietly narrowed. If it changes, we will change the number on the pricing page rather than throttle the output.

Included regardless of plan: full-resolution downloads, no watermark, one credit per restoration whichever strength you choose, and an automatic instant refund of the credit if a restoration fails on our side. Credits spent on uploads blocked by the content-safety classifier are not refunded — written into our terms and acceptable use policy, because PhotoUnfade is built for family photographs.

What resolution do you actually get, and can you print an 8×10?

This is the hidden cost almost nobody discloses, and the one most likely to make you pay twice. Our output is capped at 2,048 pixels on the long edge — a limit of the image model we use, not a paywall. Measured across our 30 completed jobs, average output was 2.42 megapixels: 17 came back at exactly 2,048 pixels on the long edge and 13 came back smaller because the input scan was smaller.

Print shops work in pixels per inch (ppi — how many pixels land in each inch of paper; 300 ppi is the standard for a print you hold in your hands). That gives a fixed pixel requirement per size:

To be plain: an 8×10 from our output looks right framed on a wall or viewed at arm's length, where 200 ppi is perfectly acceptable, but it is below the 300 ppi standard for a print people pass around a table. If a large hand-held print is the goal, that is a job for a service built around high-resolution output or a retoucher working from your master scan — better to know before spending than after. Our input limits, for completeness: JPEG, PNG or WebP up to 9 MB, long edge downscaled to 3,072 pixels if larger, short edge at least 384 pixels. The same ceiling applies to every AI restoration service, so ask any vendor before you buy: what pixel dimensions will the file I download actually be?

What hidden costs should I watch for?

Six of these have cost real families real money, and none appear in the advertised price:

• Resolution paywalls. A free result rendered at 500 pixels wide is a preview, not a deliverable. Confirm the download dimensions before paying, not after.
• Watermarks you pay a second time to remove. The output you approved and the output you receive should be the same file.
• Weekly billing dressed up as a free trial. The most common complaint against the large consumer photo apps is a trial that converts to a recurring weekly charge people cannot find how to cancel. Check the billing period and the cancellation path before the price.
• Per-download or re-export charges. You should be able to re-download your own file if you lose it.
• Postage and insurance on originals. If you mail a print, insure it — and apply the only rule that matters: if losing it would be unbearable, do not put it in the post.
• Re-scanning. The most expensive line item on any project is digitizing twice, because the first pass was at 300 dpi with the scanner's auto-enhance on.

When is it worth paying a person instead of an AI?

When the photograph is one of a kind and the damage covers a face. That combination is the whole test. AI is the right first tool for fading, yellowing, scratches, stains, small tears and general aging — the great majority of a shoebox — at cents per photo and in seconds. A human retoucher is right for the handful with major facial loss, for a reason that is not about skill but about evidence: a professional can reconstruct a specific face from other real photographs of the same person, while any model, ours included, can only produce a plausible average face from its training data. When roughly 30 percent or more of a face is gone, what appears in that gap is invention, not restoration — however attractive it looks.

There is a third answer vendors avoid naming: sometimes neither route can help. If a face is largely gone and no other photograph of that person survives, the honest options are to leave the loss visible or to accept that the result is an interpretation. Our AI versus professional comparison works through that decision. And if your worry is whether AI will quietly change a face that is still intact, we measured exactly that: skin-hue deviation of 3.4 degrees against a natural undamaged original (pass threshold under 5 degrees) and 0.0 percent invented color on black-and-white inputs — while one popular all-in-one editor put invented color into 77.5 percent of the pixels of a black-and-white family scan and produced roughly 8 times more strong-edge points than the image could physically support. Every number, including two failures we found and reported in our own product, is in our pixel-by-pixel face accuracy test.

How do I avoid paying for the same photo twice?

Work in this order, because most wasted money in this category comes from doing it backwards: digitize once, properly (600 dpi, TIFF master, every automatic enhancement off — the cheapest insurance in the project); test free on your hardest photo, not your easiest (any tool flatters a lightly faded snapshot, and it costs nothing to find out); judge at 100 percent zoom on the face, not on the thumbnail or the background, because skin tone, eyes and teeth are where over-processing shows first; and send a professional both the original scan and the AI pass, since pre-cleaning an image materially cuts a retoucher's hours and many quote lower for a pre-restored input. That last step is the difference between paying a person for the hard 10 percent of judgment and paying them for the tedious 90 percent.

The bottom line

Budget cents per photo for the volume and dollars per photo for the irreplaceable few. For a 200-print shoebox that works out to roughly $29 for the box plus a few hundred dollars for the handful that need a human — against $5,000 to $15,000 if every photo goes to a retoucher. Before you spend anything, confirm the output pixel dimensions, check the billing period and the cancellation path, and scan once at 600 dpi. To price your own box on your own photographs rather than on our averages: three free restorations, no account, no credit card, no watermark, full resolution, with the original always beside the result, and a 7-day no-questions refund on any first purchase. For choosing between tool categories rather than prices, see our honest buyer's guide; for the full scan-to-finish workflow, start with our step-by-step restoration guide.