Photos and screenshots
A document shot on a phone, a screenshot of a chat, an image from a messenger
JPGPNGHEIC from iPhoneScreenshot
Netrun OCR
Extract text from photos, scans and PDFs
Photograph a document or upload a scan you already have — a minute later you get text you can copy. Free, no sign-up
Right for most files
Need to run your own site, bot or any other project?
Netrun will host it on its servers, with no infrastructure to set up
Any document you already have
The file extension does not matter — the service looks at what is inside
PDFs and scans
Contracts, invoices, certificates, diplomas — including scans where text cannot be selected
PDFMulti-pageScan with no text
Tables
Invoices, price lists, reports: rows and columns do not get mixed together
XLSXCSVTable from a scan
Word, Excel, PowerPoint
Office files are read without Office itself — nothing to install and nothing to buy
DOCXXLSXPPTXODTRTF
Video and screen recordings
Text pulled from frames: webinar slides, captions, on-screen interfaces. Repeats are merged automatically
MP4MOVScreen recording
Archives
A whole folder of documents in one file — every one is recognised and listed in order. An archive inside an archive is unpacked too
ZIPRAR7ZTAR.GZWhole folder
Three steps and no settings
Settings exist, but only if you want them
- 1
Upload a file
Drop it on the form or click it. Several at once is fine — no sign-up, no payment
- 2
Wait a minute
The page is straightened, lighting is evened out, stamps are lifted off the text — and only then does reading begin
- 3
Take the text
Copy all of it with one button, or walk the lines: hover one and see exactly where it came from
Why it comes out more accurate
A typical online service pushes your image through one program and returns whatever comes out. Here there is more work — and it shows in the result
Every document is read twice
Two independent systems parse each page. A line missed by one is usually caught by the other — the more confident reading is the one that reaches your text
Photographs get straightened out
A shot taken at an angle, upside down, with a hand shadow and lamp glare is put right before reading. Blue and red stamps are lifted, and the text beneath them comes through
Doubtful spots are flagged
Every line carries a confidence percentage. Yellow and red lines are worth a glance — you will not have to proofread the whole document
Your documents stay yours
Files are visible only to you; someone else’s link will not open your job. Nothing goes to third-party clouds or neural networks, and you can delete everything right after downloading
The short answers
What does it cost?
Nothing. The service is free and works without registration: no card number, no email, no subscription. There are no watermarks on the result either
Who can see my documents?
Only you. The file list and results are tied to your browser — someone else’s link will not open them. Files are kept for a limited time, and you can delete any of them immediately with the button in the list
Which languages are recognised?
Russian and English, including both mixed in a single document. There is nothing to switch
What about handwriting?
Neat block capitals — a form filled in by hand, for example — usually come through. Joined-up cursive does not: those lines are marked with low confidence so you spot them straight away
Which files can I upload?
Photos and screenshots (including HEIC from an iPhone), PDFs, Word, Excel and PowerPoint documents, CSV tables, video and ZIP, RAR, 7z, tar.gz archives. The extension does not matter — the type is detected from the file itself
How accurate is it?
On a flat scan of clean print, almost flawless. On a photo taken at an angle, with glare or tiny type, mistakes are possible — which is why every line shows a confidence percentage: doubtful spots are visible immediately, not after proofreading
Do I need to install anything?
No, everything runs in the browser, on desktop and phone alike. If you need a closed environment, the service deploys to your own server with a single Docker command and then runs with no internet at all
What happens to tables?
The grid is reconstructed separately from the text, so rows and columns do not drift and merged cells are preserved. For XLSX and CSV no recognition runs at all — the data is read directly
Can I get the result as JSON?
Yes. Besides plain text you get the page structure: blocks in reading order, coordinates for every line, confidence values and cell-level tables. Useful when the result feeds your own program or a language model
Is there a file size limit?
Yes, and it is shown right on the upload form since it depends on server settings. Multi-page PDFs are processed page by page in parallel, so a long document does not crawl through single-file
Is the original layout preserved?
Reading order is preserved, including columns and book spreads. Text is returned both “as on the page”, with indentation and blank lines, and as a plain continuous flow — take whichever suits you
Photograph the page — and take the text
A minute to recognise, no sign-up and no payment