Image Batch image to text
Read text from a whole folder of images or a long PDF in one pass, then take it as one file or one text file per page.
- English
3.0 MB downloaded once, then kept on your device.
Picking the right one matters more than any other setting.
Change this if the words come back in the wrong order.
Output
Straightens a tilted page and evens out shadows before reading. Automatic applies it to photos and leaves screenshots alone.
Drop an image or PDF here, paste one with Ctrl+V, or
Add as many files as you like. Every page is read on your device, one after another.
Ready. Runs locally on your device.
Your files stay on your device. The tool works directly in your browser, using your device to process your files. Nothing is sent to our servers, and we never receive, store, or see your files or figures.
Share result
Cite
Cite this page
Reembun. (2026, July 31). Batch image to text. https://reembun.com/batch-image-to-text
How to use it
Add every file at once
Drop images and PDFs together, paste from the clipboard, or use Choose files. A PDF is split into its pages automatically.
Set the language and layout
One language setting covers the whole queue, so group your files by language before you start.
Run the queue
Each page is read on your device, one after another, with progress shown per file. You can stop partway and keep everything already done.
Take the text away
Copy a single result, use All text for the whole run, download one Word file covering every page, or Download all as ZIP to get one text file per image.
What this is for
Single-image OCR is a copy and paste job. Batch OCR is a different task with different failure modes, and it is usually one of these:
- A folder of photographed pages from a book or a file of documents.
- A scanned PDF that arrived as images and needs to become searchable.
- A month of receipts photographed one at a time.
- Screenshots of a system that has no export.
All four share the same shape: the work is boring, there is a lot of it, and losing it halfway through is worse than it being slow.
How the queue behaves
Files are added by dropping, browsing, pasting or photographing, and they accumulate rather than replacing what is already there. PDFs expand into one entry per page, so a 40-page scan becomes 40 entries you can see and remove individually.
Reading starts when you press the button, one page at a time, and each result appears as it lands. You can stop the run at any point and everything already read is kept. Pressing read again picks up only what is still waiting.
The engine and the language model download once, on the first page. Every page after that starts immediately, which is why a long batch costs much less per page than the first one suggests.
Choosing the mode before you start
Simple mode returns each page’s text with its own line breaks. It is faster, because the engine can skip building the geometry, and on a large batch that difference is real.
Formatted mode rebuilds headings, lists and tables from the layout of each page. Use it when the pages have structure worth keeping, and accept the extra time.
The mode applies to the whole queue and takes effect on the next run, so decide before you press read rather than after.
Naming, and why it matters at this scale
Every page keeps its source name, and PDF pages carry their page number. Inside the ZIP those become filenames, with characters that are illegal on Windows replaced and duplicates numbered rather than silently overwritten.
That last part is worth stating plainly. A batch tool that writes two files to the same name loses one of them, and you will not notice until you need it.
The honest limits
Accuracy is per page, not per batch. A confidence figure is shown for each page and averaged across the run. A single bad photograph in a good batch drags the average down without telling you which page it was, so read the per-page figures when the average looks wrong.
Handwriting still does not work. Doing it four hundred times does not change that.
Very large images cost memory. Anything above about 40 megapixels is refused rather than allowed to crash the tab. Phone photographs are well under that; stitched panoramas and some scanner output are not.
Why in the browser rather than on a server
A batch is exactly the case where uploading is least acceptable and most tempting. It is a lot of files, they are usually related, and together they say much more about you than any one of them does: a folder of payslips, a client’s whole contract file, a year of medical letters.
None of it is transmitted here. The engine is downloaded to your machine once and every page is read where it already is, which is also why the tool works on a plane with the connection off.
Common questions
How many files can I queue?
There is no limit written into the tool, because the limit is your device rather than our server. A modern laptop handles several hundred pages without complaint. A phone with little free memory will get slower as the queue grows, so split a very large job into two or three runs if you notice that happening.
Why does it read one page at a time instead of all at once?
Each parallel job would carry its own copy of the recognition engine, roughly 4 MB of it, plus the page it is working on. Two or three of those at once is how a browser tab gets killed halfway through a long job on a phone. Sequential is slower and it finishes, which on a batch tool is the only quality that matters.
What happens if one file fails?
It is marked as failed with the reason next to it and the queue carries on. When the run ends you can fix or remove that one file and press read again, which only picks up the pages that have not been done. A single bad file never costs you the whole batch.
What is in the ZIP?
One .txt file per page, named after the source, plus all-text.md containing everything in order with a heading above each page. Both are included because they answer different needs. Per-page files suit anything you are going to process further, and the combined file suits reading or searching.
Can I mix images and PDFs in one queue?
Yes. PDFs are expanded into their pages automatically and take their place in the queue alongside the images. Any PDF page that already contains real text is taken exactly rather than recognised, and those pages are both instant and error free.
Last reviewed
