A new way to proof, built for the AI era

Proofs your clients can see, but can't steal.

Watermarks no longer protect your proofs. Free AI tools erase them in seconds, and every clean copy is a print order you don't get. ProofProtector works differently: instead of covering your photo with a mark, it never sends the photo at all. Clients see a proof that looks right on screen, with faces marked, detail capped and a keyed shield woven in.

No client sign-up. JPEG, PNG and WebP.

Sisters · Autumn sessionClient gallery · 24 proofs
Protected
A ProofProtector proof of two sisters, with their eyes and mouths marked and the spectral shield added Real proof · what clients see
Faces marked2Auto-detected
Detail served336pxNever the original
Loupe limit60%Per viewer
Spectral shieldKeyed noise shaped to each photo's own detail, to trip up AI clean-up. Faces markedEyes and mouths carry a PROOF mark, automatically. Metered loupeClients check focus up close, up to 60% of each photo.

Built for photographers who sell the print

  • School & nursery
  • Weddings
  • Newborn
  • Family portraits
  • Sports teams
  • Events
  • Headshots

The problem

Watermarks no longer protect your proofs.

AI made removing a watermark a one-tap job. Your clients don't have to be thieves: when a clean copy is free and instant, plenty of people take it instead of ordering. Every one is a print sale you lose.

  1. ThenHours

    Photoshop and patience

    Removing a watermark took skill, paid software and an afternoon. Most people didn't bother, so a visible mark was enough.

  2. RecentlySeconds

    Free AI removers

    Upload a proof to a free website and the mark is painted out in seconds. No skill, no software, no cost.

  3. NowOne tap

    Built into phones and chatbots

    Phone galleries ship one-tap erase tools, and chatbots will redraw a photo from a screenshot on request. Each new model makes it easier.

Why the usual fixes stopped working

  • Visible watermarks

    A mark repeated across a gallery can be stripped automatically1, and in 2025 people removed Getty Images watermarks with a single Gemini prompt2.

    95% of faces still recognisable after an AI remover, in our tests.

  • Right-click and download blocking

    The full image is already on the client's screen. A screenshot captures it in one press, no tools needed.

    100% of the image captured by a screenshot.

  • Small, low-res proofs

    Free AI upscalers sharpen a small proof back up, and the faces, the part that matters, come through intact.

    82% of faces still recognisable after an AI upscaler, in our tests.

  1. Dekel, Rubinstein, Liu and Freeman (Google Research), On the Effectiveness of Visible Watermarks, CVPR 2017.
  2. TechCrunch, People are using Google's new AI model to remove watermarks from images, March 2025.

What it costs you

A few lost orders a month adds up to thousands a year.

Most clients who keep a free copy never tell you. You just see a smaller order, or none. Move the sliders to match your business.

Print sales lost each year £3,000

An estimate from your own numbers, not an industry figure.

The old way

Cover the photo with a mark and hope nobody removes it. The full photo still reaches the client, so any AI tool has everything it needs.

The ProofProtector way

Never send the photo. Clients get a proof with capped detail, marked faces and a keyed shield. There's no clean copy for any AI to recover, because it never left the server.

See the AI test
Press and hold to inspect
Loading the sample gallery…
Loading…

Live sample

Press and hold. Inspect it like your client would.

This is the real ProofProtector viewer, running on our sample gallery. Hold down on the photo (or touch and drag on a phone) to open the loupe and check focus and expressions up close.

  • Take a screenshotYou'll capture a 336px proof with marked faces and the shield woven in, not the photo.
  • Keep inspectingThe loupe is built from small marked tiles. After 60% of a photo, it stops.
  • Look closerEvery loupe tile carries a faint PROOF mark, and the faces outside the loupe stay marked.
Try it with your own photo

Features

Everything a proof needs. Nothing a thief can use.

Four layers of protection that work together, so you can share proofs freely and still sell the prints.

Spectral shield

Every proof carries keyed, mid-frequency noise shaped to the photo's own detail and hidden in its texture. It's the same for every copy, so it can't be averaged away.

Faces marked automatically

Face detection finds every face and burns a PROOF mark over the eyes and mouth, exactly where a stolen photo would need to be clean.

Press-and-hold loupe

Clients can check focus and expressions up close. Each viewer may reveal at most 60% of a photo, in small marked tiles, rate limited.

Private by default

Originals sit in private storage and are never served. Proof images are fetched through signed links that expire within minutes.

One link for your client.
Nothing worth stealing.

Share a gallery link by email or text. Your client opens it in any browser, no account needed.

Open the sample gallery

How it works

From shoot to shared gallery, in seconds.

Upload as usual. ProofProtector builds every protected file on the server, so the browser never holds a clean copy of your photo.

01Upload

Uploading 24 photos

  • Resized in your browser
  • Signed private upload
  • Original stored privately
  • Proofs rendering

Upload your shoot.

Your full-size originals go straight to private storage, with a 2400px copy for rendering proofs.

02Protect

Proof renderer

  • Detail capped at 336px
  • Faces found and marked
  • Spectral shield added
  • Marked loupe tiles

We build the proofs.

The renderer runs on the server and never leaves it. Each photo is ready in seconds.

03Share

Client link

Loupe use, this viewer34% of 60%

Send one link.

Clients browse, press and hold to inspect, and pick favourites. You keep the originals.

Why it holds up

Watermarks get removed. Missing detail doesn't come back.

Visible watermarks and right-click blocking stop casual saving. They don't stop a screenshot or an AI cleanup tool.

The original photo: two smiling sisters
OriginalYour photoThe full-quality image from the shoot. It's never sent to your client.
The watermarked proof after an AI watermark remover: clean and almost identical to the original The proof that went in, with a tiled PROOF watermarkBefore
Watermark + AI removerA near-perfect copyA free AI remover erased a standard tiled watermark in seconds. The faces and the detail came back.
The ProofProtector proof after the same AI remover, with the mark areas marked for removal: soft, with the faces not restored The ProofProtector proof that went in: 336px of detail, shielded and marked over the eyes and mouthsBefore
ProofProtector + AI removerNeither face recognisableThe remover never saw the eyes or mouths, so it had to guess them, and it only had 336px of detail to work from. Your original never left the server.

Real output, not a mock-up. Middle: LaMa, the open-source AI model behind many free watermark removers, with the exact watermark area marked for removal. Right: a ProofProtector proof through the same model, with its marks marked for removal. The ProofProtector proof includes the spectral shield.

Faces an AI can still recognise

Share of faces that face-recognition software still matches to the real person. Lower is better.

9%

of faces recognisable in a saved copy of a ProofProtector proof

Visible watermark: 100%

5%

recognisable after an AI watermark remover

Tiled watermark: 95%

41%

of the original's image quality left after the remover

Tiled watermark: 99%

As capturedScreenshot or save, no clean-up
  • ProofProtector9%
  • Tiled watermark100%
  • Corner watermark100%
  • Low-res JPEG95%
After an AI watermark removerLaMa, with the mark area marked
  • ProofProtector5%
  • Tiled watermark95%
  • Corner watermark100%
  • Low-res JPEGNot tested
After an AI upscalerReal-ESRGAN, 4×
  • ProofProtector14%
  • Tiled watermark100%
  • Corner watermark100%
  • Low-res JPEG82%

Measured on our 6 sample photos (22 faces), so treat the figures as early results. Faces are matched with OpenCV's SFace model at a threshold only 1% of pairs of different people reach; image quality is MUSIQ, as a share of the original's score. ProofProtector figures are for the proof as served, spectral shield included. With only 22 faces, the shield's effect on face matching is within the margin of error; its clearest effect is on quality, which the shield lowers from 55% to 41% after the remover.

ProtectionVisible watermarkRight-click blockingProofProtector
ScreenshotClean copy, mark includedFull image captured336px, faces marked, shielded
AI watermark removalOften removed in secondsNothing to removeOnly 336px of detail to work with
Original in the browserUsually full sizeYes, in the pageNever sent
Faces usable in a cropIf the mark misses themYesEyes and mouths marked
Client can check detailYesYesMetered loupe, up to 60%

Pricing

Plans that grow with your bookings.

Start free with a full gallery. Upgrade when you're ready to send proofs to every client.

Monthly Yearly Save 20%
01Starter

Starter

Try ProofProtector on a real shoot.

£0/mo

One gallery, free for as long as you like.

  • 1 active gallery
  • Up to 25 photos
  • Spectral shield and face marking
  • Metered loupe
Start free
02Most popular

Studio

For working photographers sending proofs every week.

£12/mo

Everything you need to proof every client.

GalleriesUnlimited
Photos5,000per month
  • Unlimited galleries
  • 5,000 photos a month
  • Your logo on client galleries
  • Per-gallery loupe limits
  • Email support
Start with Studio
03Pro

Pro

For schools, events and multi-photographer studios.

£29/mo

High volume, team seats and priority help.

  • Everything in Studio
  • 25,000 photos a month
  • Team seats
  • Custom domain for galleries
  • Priority support
Talk to us

No setup fees. No contracts. Cancel any time.

FAQ

Questions, answered.

What photographers usually ask before sending their first protected gallery.

Can clients still take a screenshot?

They can, but all they get is the proof: 336px of real detail, marked faces and the spectral shield woven through it. That's fine for choosing favourites and no use for printing.

Will my clients notice the shield?

Up close it shows as a fine grain, mostly in textured areas like hair and foliage; flat areas like sky are kept cleaner. At normal viewing size the proof reads as your photo.

Does it stop AI watermark removers?

The mark is burned in after the detail is cut, the shield is added on top, and the original is never sent. Removing the mark would still leave a soft 336px image, which is no use for printing. The shield isn't tuned against any particular AI model, so we don't claim it defeats them on its own.

Won't AI tools keep getting better?

Yes, and that's why we don't rely on a watermark. A better model can only work from what it's given: a 336px proof with marked faces. It can guess at what's missing, but the guess isn't your photo and isn't fit to print.

What happens to my original photos?

They're kept at full size in private storage. Proofs are rendered on the server from a 2400px copy made in your browser. Neither is ever served to a viewer.

How closely can clients inspect a photo?

Pressing and holding opens a loupe built from small, marked tiles. Each viewer can reveal at most 60% of any one photo, and requests are rate limited.

Do clients need an account?

No. You send a gallery link and they open it in any modern browser, on phone or desktop.

Ready when you are

Protect your next gallery with ProofProtector.

See a protected proof in seconds. Upload one of your own photos and see what your clients get.

  • Originals never served
  • No client sign-up
  • Ready in seconds