OCR vs. handwritten text recognition
Traditional OCR engines such as Tesseract match shapes against known fonts. That works for a scanned book but not for handwriting, where every writer forms letters differently. Handwritten text recognition (HTR, sometimes called ICR) reads whole words in context, the way a person does.
This tool takes the HTR approach for everything, so a page that mixes typed questions with handwritten answers comes out as one readable transcript.
| Classic OCR | This handwriting OCR |
|---|---|
| Best on clean printed fonts | Reads print and handwriting |
| Character-by-character matching | Reads words in context |
| Struggles with cursive and slanted lines | Copes with cursive, skewed photos and ruled paper |
| Often needs you to pick a language pack | Detects the language automatically |
Typical jobs people use it for
- Typing up filled-in paper forms and questionnaires
- Pulling margin notes off a printed contract or draft
- Transcribing family letters and postcards for genealogy
- Digitising lab notebooks and field notes
- Getting text out of a photographed whiteboard after a meeting
Tips for higher OCR accuracy
- Crop out your desk, hands and anything that isn't the page
- Avoid shadows across the writing; tilt the page toward the light
- One page per photo gives better results than a whole spread
- Pencil reads better when the photo is slightly overexposed
- If one word keeps coming out wrong, a closer photo of just that part usually fixes it
Mostly handwritten pages? Try Handwriting to text converter.
Frequently asked questions
Which languages does the handwriting OCR support?
It detects the language automatically and works with Latin, Cyrillic, Arabic, Chinese, Tamil, Devanagari and many other scripts. For the best results in a known language, use one of the language-specific OCR pages listed below.
Is it better than Google Lens for handwriting?
Lens is great for quick lookups. This tool is built for transcription: it keeps the original line breaks, marks unsure words with [?] instead of guessing silently, and lets you process several pages and download the text.
Does it work on printed text too?
Yes. Printed pages, screenshots and photos of signs all work. Handwriting is simply where it does better than older OCR tools.
Why can't I upload PDFs?
This tool is tuned for single images so each page gets the model's full attention. Screenshot or photograph PDF pages and upload them as images.
How many images can I process?
Up to five at a time, each under 10 MB. Each image uses one credit.