AI image proofreader
AI proofreader for images, ads and PDFs
Proofread the finished file — the one that is about to go live — and check every word on it against what was approved.
LYNCE is an AI image proofreader that checks text and details inside JPG, PNG, WEBP and PDF files. It detects spelling, grammar, pricing, date, layout and consistency mistakes directly on finished artwork. You can also upload a campaign brief, approved copy, price list, master creative or other reference document, and LYNCE will cross-check every file against the approved information.
It is the last check before a file goes out — the step between “design is done” and “it’s live”. If you are looking at a wider creative QA workflow across a whole campaign, that is the same product seen from further back.
Read on the artwork
€45 — the price list says €40
Fix Change the badge to €40 to match the approved price list.
“Sunday 12 July” — 12 July is a Saturday
Fix Check the date against the brief, then correct the day or the number.
What LYNCE checks on an image.
Proofreading a design is not proofreading a document: the words are short, they sit in different places, and most of them are facts rather than sentences.
Spelling and grammar
Typos, doubled words and broken grammar in headlines, body copy, buttons and small print — read in place on the artwork, in the language the creative is written in. This is the part people mean by an image spell checker; it is the floor, not the ceiling.
Prices, dates and numbers
Prices, discounts, dates, opening times, phone numbers and URLs. These are the details that cost money when they are wrong, and the ones an eye slides past after the fourth look at the same layout.
Claims and offers
What the creative promises, checked against what was approved: the offer, the terms, the disclaimer, the asterisk that should have had a footnote.
Layout and truncation
Text that is cut off, overlapping, missing from a version, or spilling out of its box after a translation ran longer than the original — the visual proofreading a text tool never sees because it never sees the design.
Consistency across versions
Scan the set together and every version is checked against the others: the €40 that is still €45 in one size, the date that changed everywhere except the story format. See ad proofreading for the campaign case.
Checked against what was actually approved.
This is the part a general-purpose proofreader cannot do. A sentence can be perfectly written and still be wrong, because the brief said something else.
What was approved
What the creative says
Upload a brief, approved copy, a price list, brand guidelines, a master creative or any other reference document, and every file in the scan is compared with it. Translated versions are checked against the master rather than proofread in isolation — see translation QA — and brand-level rules live in the brand consistency checker. Reference files are free: they don’t use credits.
How the workflow runs.
Upload the creatives
One image or the whole campaign: every size, every language version, or a multi-page PDF. JPG, PNG, WEBP and PDF.
Add references, if you have them
The brief, approved copy, a price list, the master creative, brand guidelines. Optional — and reference files don’t cost credits.
LYNCE reads and checks
It reads the text where it sits on the artwork, checks it for language mistakes, compares it with your references, and compares the versions with each other.
Work through the report
Findings are pinned to the artwork with a severity and a suggested fix. Share the report by link; no login needed to read one.
The same four steps, shown screen by screen, are on how it works.
The product, screen by screen.
Not a mock-up: this is the interface you get after signing in, walked through with sample files.
Recent scans, findings and credits for the whole workspace.
Dashboard
Creative QA board
Track recent scans and findings across your workspace.
130 credits remaining
70 of 200 used
- Total scans
- 24
- Issues found
- 57
- Clean scans
- 9
Recent scans
View all scansautumn-menu.pdf
2 hours ago
Issues found · 2Open reportThe real LYNCE interface, shown with sample files.
Formats and languages.
Image formats
JPG, PNG and WEBP — exports, not source files. Any size and aspect ratio: feed, story, banner, poster, menu board.
PDFs
Multi-page PDFs are read page by page, so page 40 gets the same attention as page 1. Brochures, catalogues, menus, decks and print-ready documents. See PDF proofreading.
Whole campaigns
Several files in one scan, checked individually and against each other. A 12-version campaign is one scan, not twelve reviews.
Languages
LYNCE reads creatives across many languages, including artwork that mixes two. We don’t publish a fixed language list, because coverage is not a switch that is on or off — if a language matters to you, the free credits are enough to test it on your own files before you pay anything.
Reference documents
PDFs and documents you already have: the brief, approved copy, price lists, guidelines. They are read for comparison and don’t consume credits.
Who proofreads images with LYNCE.
Design teams
The last check before handoff, on the export rather than the working file.
Agencies
Client work, many versions, one reviewer. Findings arrive as a shareable report with the evidence attached.
Marketing teams
Campaign sets that must all say the same thing across sizes, channels and languages.
Freelancers
Nobody else to check the final file — this is the second pair of eyes.
Not OCR, not a grammar checker, not another manual pass.
Versus OCR
OCR turns pixels into characters and stops there — you get text and no opinion about it. Reading the artwork is LYNCE’s first step, not its output: after reading, it judges what it read against language, your references and the other versions in the scan.
Versus grammar checkers
A writing assistant checks text you paste into it, and judges each sentence alone. It cannot open a finished JPG, and “Tickets €45” looks perfect to it. LYNCE proofreads the exported artwork itself and checks the facts on it. See creative proofreading.
Versus manual proofing
People are good at this for ten minutes and for the first three versions. LYNCE reads version 40 exactly as carefully as version 1 — then a human decides what to change. It reports; it never approves or blocks a file.
Versus proofing platforms
Tools like Filestage route files, collect comments and track approvals; a person still has to find the mistake. LYNCE runs before that step and hands your reviewers a list instead of a blank page. See the comparisons.
What it does not do.
The useful version of this page includes the limits. Every one of these is deliberate.
No visual brand policing
Logo placement, fonts, colour values and safe areas are not measured. LYNCE guards what the creative says, not how it is drawn.
No approvals or blocking
There is no gate, no pass/fail score and no lock on your files. Findings come with severity and a suggested fix; the decision stays with your team.
Not a guarantee
No proofreader catches everything, human or otherwise. LYNCE is another careful reader before sign-off, and it is at its best when you give it the references to check against.
Not a design editor
LYNCE doesn’t rebuild your layout or ship the fix for you. It tells you exactly what is wrong and where it is, and your designer makes the change in the tool the file came from.
None of this replaces a method. If you want the manual version — the checklist a person can run on any file, with or without software — it is in how to proofread an image.
AI image proofreading, answered.
The final check, before it goes live.
Give the finished creative one more QA pass before publish, send or print.
Related