Read the Text in a Photo or a Scan OPTIC
Every image-to-text site works by uploading the picture, and the pictures people most want read are the ones they least want copied anywhere: a passport page, a prescription, a contract photographed on a desk, a screenshot of somebody else's message. OPTIC runs the reading engine inside this tab. The text comes out, the picture stays here.
1. Add the picture or the scan
A photo, a screenshot, or a PDF that is really a scan. JPEG, PNG, WebP, GIF, BMP, HEIC or PDF, as many as you like. Drop them, choose them, or paste a screenshot.
The first time, the reading engine (about 4 MB) and a language file (about 3 MB) are fetched from this site once and kept on your device. After that it works with the internet off.
2. What it read
surefairly surecheck this one against the picture
What this is, and what it is not
The reading happens here. The engine is Tesseract, the same open source recogniser behind most scanning software, compiled to run inside this browser tab. It and the language file come from obscuraos.com the first time and are kept on your device. The picture is never sent anywhere, and you can watch the network while it works to check that.
It tells you what it is unsure of. Every word comes back with a confidence, and the doubtful ones are outlined in red on the picture. A number that reads as 1 or l, a 0 that might be an O: those are exactly the words to look at, and the box shows you where. Read the text as a draft with the doubts marked, not as gospel.
Printed text, reasonably square. Typed and printed text at a sensible size reads well, in nineteen languages. Handwriting does not, and no honest tool will tell you otherwise. A photo taken at an angle or in poor light reads worse than a scan; PHOTO can crop and straighten it first.
A searchable PDF keeps the picture. Tick the box and each page is written back as a PDF with the scan exactly as it was and an invisible layer of text on top. It looks identical, and it is now something you can search, select and copy from. Nothing in the picture is changed.
Questions people ask
Is this really not uploading my picture?
Yes. The engine runs in a worker inside this tab; the only things fetched are its own code and the language file, from this site, once. Open the browser's network panel while it reads and you will see no request carrying the image. That is also what the site's own checks watch for on every tool.
How accurate is it?
On a clear scan of printed text, most pages come back over ninety-five percent sure and read cleanly. A phone photo in mixed light, a fax, or a small font drops that, and the page confidence shown beside each page tells you which case you are in. The words it doubts are outlined on the picture so you can check them rather than guess.
Can it read handwriting?
Not usefully. It is trained on printed letterforms and will produce something for handwriting, but that something is mostly wrong. It is said plainly here because the sites that claim otherwise are selling the upload, not the result.
What languages does it read?
English, French, German, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Turkish, Swedish, Danish, Norwegian, Finnish, Czech, Arabic, Japanese and simplified Chinese. Each language is a separate file of about three megabytes, fetched the first time you choose it. A page has to be read in the language it is written in; tell it which.
What is a searchable PDF?
A scan is a picture of a page, and a picture cannot be searched. Ticking the box writes each page back as a PDF that carries the picture unchanged with the recognised text laid invisibly over it, so a PDF reader can search, select and copy the words while showing you the original scan.
Does it work offline?
After the first run, yes. The engine and the language file are kept on the device by the site's service worker, so the page reads pictures with the internet switched off. Choosing a new language needs one more download for that language.