Image spell checker
Image spell checker for finished artwork
Check spelling in a JPG or PNG where the words actually sit, and catch the mistakes a spell checker was never built to see.
LYNCE is an image spell checker for finished artwork. Upload a JPG, PNG or WEBP (a social post, a poster, a banner, a menu) and it reads the text where it sits on the image, checks the spelling and grammar, then checks what a spell checker can’t: prices and dates against your brief, versions that disagree with each other, translations against the master, and text that is cut off. Findings come back in under a minute, pinned on the image, each with a suggested fix.
Spelling is where it starts, not where it stops. The whole product (every format, every reference document, the full workflow) is on the AI image proofreader page. This one answers the narrower question: how do you check spelling in an image at all?
Free delivery on orders over €50
Read on the image
“30% of” — a real word, in the wrong place
Fix Change “of” to “off” in the headline.
€29 — the brief says coats from €24
Fix Change the badge to €24 to match the brief.
“Free delivery on orders ov…” runs off the edge
Fix Shorten the line or reduce the type so the whole sentence fits.
Why a spell checker can’t check spelling in an image.
It isn’t that spell checkers are bad at images. They never see one: they work on text, and a finished image hasn’t got any.
It’s pixels, not text
A JPG or PNG stores colours, not letters. As far as the file is concerned, the headline on your poster is the same kind of thing as the photo behind it: a pattern of pixels. There is nothing to select, nothing to paste and nothing to underline.
The export flattens the live text
In the design file the text is live: you can click into it, edit it and, if your editor offers it, spell check it as you type. Export to JPG or PNG and that layer is flattened into the picture. The file that reaches the client, the printer or the ad platform has no words in it, only the shapes of words, and any edit made after the last check goes out unchecked.
OCR gives you text, not judgement
Optical character recognition turns pixels back into characters, which is a genuinely useful first step. Then it stops: you get a block of text with no opinion about it. Stylised headlines and text set over photos come back garbled, so the spell checker flags OCR’s mistakes alongside yours, and a word that is spelt correctly but wrong, like “30% of”, sails through both.
Spelling, and everything spell check can’t see.
A spell check answers one question: is this a word? On finished artwork, that is the smallest question on the page.
Spelling and grammar
Typos, doubled words, missing words and real words in the wrong place, read in place on the image: headlines, buttons, badges, text over photography and the small print nobody enlarges. In the language the artwork is written in, including images that mix two.
Prices and dates against the brief
Add the brief, the price list or the approved copy, and every price, date and time on the image is compared with it. “€29” is a perfectly spelt price; it is still wrong if the brief says €24. Reference files don’t use credits.
Version mismatches
Upload the whole set, every size and every channel, and the versions are checked against each other. The square that got the final edit and the story format that didn’t come back as an inconsistency, with both quoted.
Translations
Translated versions are checked against the master rather than spell checked on their own: a discount condition that flipped, a date that moved, a line that was simply left out. More on translation QA.
Cut-off and overlapping text
The line cut off in the 4:5 crop, text running into a photo, a translation that no longer fits its box. A spell checker reads text, so it can never tell you the text isn’t all there.
Contact details
Phone numbers, email addresses, URLs and opening times: the details nobody owns and everybody assumes someone else checked.
How to check spelling in an image with LYNCE.
Upload the image
JPG, PNG or WEBP: one file, or a whole campaign set at once, uploaded directly or imported from Google Drive. Add the brief or price list as references if you have them. They’re optional, and they don’t use credits.
Findings in under a minute
LYNCE reads every line on the image and checks it. Each finding is pinned on the image where the mistake sits, and ranked: must fix, inconsistency or worth checking.
Fix it, then share it
Every finding comes with a suggested fix, and a person makes the final call: LYNCE reports, it never approves or blocks a file. Share the report by link with whoever needs to see it.
One image is one credit, and so is one PDF page. Plans start at €11 a month excluding VAT; see pricing.
Spell check a Canva or Photoshop export.
Check the file that goes out, not the file you made it in.
Export the design as JPG, PNG or PDF, exactly as you would for the client, the printer or the ad platform, and upload that. It is the version people will actually read, and the one where the last-minute edit either made it in or didn’t. There is no plugin to install and no source file to share, and LYNCE never changes your design: it tells you what is wrong and where, and your designer fixes it in the tool the file came from.
If you only need a quick one-off spell check.
Then you don’t need LYNCE, and we would rather say so.
For one image, once: run it through a free OCR tool to get the text out, paste the text into an ordinary spell checker, and read the result against the picture. OCR misreads stylised type, so some of what gets flagged will be its mistakes rather than yours. Then check every price, date and phone number against the source by hand. That is a perfectly good basic check, and the full manual method is in how to proofread an image.
LYNCE is for people who check finished artwork regularly: in-house teams, agencies, studios sending files to print, anyone running campaigns across sizes and languages. When it is forty images a week, the price list changed on Tuesday and the German version has to say what the English one says, copying text out of pictures stops being a method and starts being a job.
Image spell checking, answered.
The final check, before it goes live.
Give the finished creative one more QA pass before publish, send or print.
Related