What photo restoration actually costs.

Seven ways to get a photograph restored, and what each really costs.

These are order-of-magnitude ranges for Western markets in 2026 and vary significantly by country. A studio in a low-cost region can be excellent and a fraction of the price; the range is a sanity check, not a quote.

Disclosure: Pinin is run by the same team that publishes this site. Its row is on this table because leaving it out would make the comparison less useful, not more. The figures are its published rates, checked in October 2026 — and most of the other rows here exist to talk you out of buying it.

Repair is cheap. Reconstruction is not. Everything else is detail.

This is the single most useful thing on this page. Restoration work splits into two jobs that look similar in a portfolio and are not remotely the same:

Repair is removing and correcting things that are there but should not be: scratches, dust, foxing spots, creases, a colour cast, fading, blur, noise. Every pixel of the original subject survives. The work is subtraction, and subtraction is exactly what software is good at. This is why an automated service at a dollar a photo can genuinely match a competent human retoucher on a faded 1970s print.

Reconstruction is inventing things that are not there: a torn corner with half a face, a water stain through someone's shoulder, a hole where the emulsion lifted. There is no information to recover, so the retoucher has to decide what the missing area looked like. That is judgement, and judgement is what you pay a human for.

So the question is never "is AI good enough?" It is "how much of this photograph is missing?" Fading, scratches and colour shift: automation wins on price and usually on quality. Missing features: pay a person, and ask them what they invented.

What actually moves the price.

• Whether faces are involved. Skin, eyes and teeth are where invented detail becomes obvious. A damaged landscape is cheap to fix; a damaged face is not.
• Missing area, not damage area. A hundred scratches is repair. One missing square centimetre containing an eye is reconstruction.
• Original versus copy. Restoring from a print is straightforward. Restoring from a photograph of a print taken on a phone is a different and much worse starting point — see the triage tool for why that matters.
• Resolution and intended output. A file for sharing at screen size needs a fraction of the work of a file destined for a 24-inch framed print.
• Colour. Colourising black-and-white is a creative interpretation and is priced as such. Correcting a colour cast is mechanical and cheap.
• Volume. Studios quote per batch more cheaply than per photo. A hundred photos from one collection is a different job from one photograph.
• Rights and licensing. If you are restoring for commercial use, expect a different conversation about who owns the result.

The cases where the cheaper route is a false economy.

1. The photograph is the only surviving image of someone, and part of them is missing. Reconstruction judgement is worth paying for once, and this is the once.
2. It will be printed large and framed. Artifacts invisible on a phone are obvious at A2.
3. It has archival, legal or estate significance. An invented detail in a probate or insurance context is a problem, not a feature.
4. The original is itself valuable. Spend accordingly on the file that represents it.
5. The damage is a composite of several types plus missing areas, where each fix affects the next.

When you hire, ask three questions before paying: may I see two examples of similar damage?, will you send the unretouched scan as well as the restored file?, and which parts of this image did you invent? A good restorer answers all three without flinching. Insist on receiving the untouched scan — it is your insurance against a restoration you dislike.

The cases where paying a human is just paying more.

• Fading and colour shift — the classic 1970s print that has gone pink or cyan.
• Scratches, dust, speckling, foxing — subtraction work, where automation is strong.
• Blur and softness, within reason. Beyond reason is invention, and it shows.
• Colourising black-and-white, where the honest framing is "a plausible interpretation" rather than "the true colours".
• Batches. Fifty family photographs at studio rates is thousands; the same fifty done automatically is coffee money, and for these damage types the output is comparable.

The sensible default for most family collections: run everything through an automated service first, because it is nearly free and fixes the majority of cases. Then take the two or three photographs where the result was not good enough — usually the ones with something missing — to a human.

Pinin runs on that model: your first restoration is free, then you pay per use rather than per subscription — the pricing is here — so trialling it on your worst photograph costs nothing.

The costs nobody quotes you.

• Scanning. Often the largest real cost and the one people forget. A restoration needs a good source file first — here is what scanning costs.
• Posting the original. Irreplaceable prints in the mail means insurance, tracking, and a real risk. Ask whether you can work from a high-resolution scan you send digitally. Most can.
• Printing and framing. Frequently more than the restoration itself, and it is where the file finally gets judged.
• Storage and backup. A restored file that exists in one place is one dead laptop from gone. The 3-2-1 rule applies here too.
• The second restoration. If you accept an over-smoothed, over-brightened result, you may end up paying twice. Keep the original scan.