i2OCR
Pulls text out of an image or a scanned PDF in 128 languages and scripts, free and with no account.
Overview
i2OCR reads the words in a picture. Give it a photo of a page, a scan or an image-only PDF, and it returns text you can edit or search. It covers 128 languages and scripts, and you pick the language yourself. You also set single or multi-column reading order. The output can be plain text, Word, HTML, or a searchable PDF that looks like your scan with the text invisibly behind it.
It runs in the browser. The free tool needs no account and takes one image or one PDF page per run, with a paid bulk option. Uploads go up to 20 MB for an image and 200 MB for a PDF. The vendor states uploaded files and the extracted text are deleted automatically after processing, currently within about half an hour.
Key features
When to use
Use i2OCR if:
- You have a scan or a photo of a document and you want the words out of it.
- Your document is not in English. It covers 128 languages and scripts, including Arabic, Cyrillic, Asian and Indic writing.
- Your page has columns. You can tell it to read multiple columns rather than straight across.
- You want the scan to stay a scan but become searchable. The searchable PDF output does exactly that.
- You want it now, free, with no account and nothing installed.
- You have a PDF that is a picture rather than text, so copying from it does nothing.
When NOT to use
Skip i2OCR if any of these describe you:
- You have many pages. The free tool takes one image or one PDF page per run, so a long document means repeating it, and bulk work is the paid option. Compare other image to text tools.
- Your page is a table you want as a spreadsheet. OnlineOCR exports to Excel with the layout kept.
- Your source is handwriting and you need it accurate. i2OCR states it does recognise handwriting, but less accurately than typed text, and Pen2txt targets handwritten notes.
- Your scan is blurry, skewed or low resolution. It expects a reasonably sharp image, and a poor scan produces poor text whatever the tool.
- The document is confidential. It is uploaded to be read, and the vendor states files are deleted after processing rather than never stored.
- You expect a perfect transcript. Every OCR misreads characters, so the output needs checking.
Frequently asked questions
The web tool is free and needs no account or installation. It works on one image or one PDF page per run, and the vendor puts no daily limit on how many runs you make. A paid bulk option exists for larger workloads. A long document means paying, since the free tool takes one page per run.
It is your scan, unchanged to look at, with the recognised text stored invisibly behind the image. You still see the original page, but you can search it, select text and copy from it. That suits archiving a document where the original appearance matters, rather than pulling the words into a new file.
Because it decides the reading order. On a page with two or three columns, a tool set to single column reads straight across and mixes sentences from different columns together. Telling it there are multiple columns makes it read down each one in turn, which is how the page was meant to be read.
Good enough to save the typing, not good enough to skip checking. OCR recognises shapes, so it confuses similar characters, and receipts, unusual fonts and poor scans make that worse. Read the output against the original, especially for numbers and names, where a mistake will not look wrong.
A sharp, straight, well-lit one at a decent resolution. It takes JPG, PNG, TIF, BMP, GIF and WEBP images up to 20 MB, and PDFs up to 200 MB. A photo of a page taken at an angle in poor light gives the recognition much less to work with. If the scan is skewed, straightening it first will improve the result more than trying a different tool.
It is uploaded to be processed. The vendor states that uploaded files and the extracted text are automatically deleted after processing, currently within about half an hour. For a confidential document, note that the file does reach its servers.
Alternatives to i2OCR
Image to Text
Free OCR that pulls the text out of an image, built on the open-source Tesseract engine, with no signup.