How to read a scanned book or handwritten page in another language

You found the perfect study material — a chapter from a borrowed textbook, a photocopied handout from class, a page of a novel you snapped in a bookshop. There's just one problem: it's an image. You can't tap a word to look it up, you can't copy a sentence into a dictionary, and you certainly can't hear it read aloud. So you end up retyping words letter by letter into a translator, losing your place every few seconds.

There's a much faster way. With OCR — optical character recognition — you can turn a scan or a photo into real, selectable text. From there you highlight any word to translate it in context and hear it pronounced. Here's how to do it well, and how to get recognition clean enough to actually study from.

First, know what kind of file you have

There are two kinds of PDFs, and the difference decides everything:

• Born-digital PDFs were created from a document — exported from a word processor, a publisher's ebook, an article you downloaded. The text is already "live," so you can select it. In Glotsmith a page like this reads for free, with no limits, because there's nothing to recognize. (A thin page — a title page, or one that's mostly a diagram — can still fall back to OCR.)
• Scanned or photographed pages are pictures of a page — a scan of a printed book, a phone photo, an image file. The words are just pixels until OCR reads them. Glotsmith runs OCR on these, and OCR is metered on every plan: you get a monthly allowance of pages, and the paid plans raise it.

Quick test: open the file in any reader and try to drag your cursor across a sentence. If the words highlight, it's born-digital. If your cursor just draws a box over a flat picture, it's a scan and needs OCR.

When you have a choice — a downloadable ebook versus a scan of the same book — pick the born-digital version. It's cleaner, it's free, and there's nothing to wait on.

Get the cleanest scan or photo

OCR is only as good as the image you feed it. A few habits make the difference between crisp text and a garbled mess:

• Flatten the page. Curl and the shadow near a book's spine confuse recognition. Press the book flat, or use a scanner's lid.
• Light it evenly. Bright, diffuse light beats a single lamp. Watch for glare on glossy pages and for your own shadow falling across the text.
• Shoot straight on. Hold the camera parallel to the page and fill the frame. Skewed, angled text lowers accuracy fast.
• One page per image. Don't cram a two-page spread into a single shot; crop tight to the text you want.
• Favor contrast and resolution. Sharp black type on white reads best. If your scanner has a "text" or "document" mode, use it.
• Skip the marked-up copy. Heavy highlighting, underlines, and scribbled notes over the words can trip recognition. A clean copy wins.

From image to text you can study, step by step

Say you've photographed a page from an Italian textbook. Here's the flow:

1. Create a course for the language you're studying, then add a workbook for this text.
2. Upload the file as a document — a PDF, or a photo of the page as a JPG, PNG, or WebP. (Many phones shoot in HEIC by default; export or convert to JPG first.)
3. Let it read the page. A born-digital PDF opens instantly; a scan or photo goes through OCR, which turns the pixels into selectable words.
4. Highlight to translate. Drag across a word or a whole phrase to get an in-context translation. Grab the full phrase, not just the single word — context is exactly what a plain dictionary can't give you.
5. Hear it read aloud. Play your selection in a natural voice in the source language. It's the fastest way to fix a pronunciation you've only ever seen on the page.
6. Save the keepers. Words worth remembering collect in one course-wide vocabulary list, each tagged with the workbook it came from, so you always know where you met it.

Prefer to work line by line?

For a dense paragraph you want to unpack slowly, use the side-by-side translation editor: put the passage in and translate it row by row, source and target visible together.

What about handwriting?

Be honest with yourself here. OCR was built for printed text. Neat, tidy printing — block letters, a careful hand — often comes through well enough to study from. Flowing cursive, messy, or stylized writing is unreliable; you may get a partial reading, or a scrambled one. Photograph it as cleanly as you can, then treat the result as a starting point: highlight the parts that came through, and read the image directly for the rest. Printed scripts, including Japanese and Korean type, are the sweet spot; a handwritten letter in loopy cursive, much less so.

Turn one page into a real study session

Once the text is live, the page becomes more than something to translate. Pin your own notes to a spot, build a small table of the conjugations that keep tripping you, add explanations in your own words. If you also have a recording you've uploaded — the chapter's audio, or a podcast episode on the same topic — generate a timestamped transcript and highlight lines there to translate them too (that method has its own guide). When you want a tutor's eyes on it, share a read-only link.

One method tip worth more than any tool: read the whole page once for the gist without stopping, then go back and highlight only the words that actually blocked your understanding. You'll spend less of your monthly translation budget and read the way you already do in your own language — for meaning first.

Start with the page in front of you

A scanned chapter or a photographed page doesn't have to stay locked as an image. Feed it a clean scan, let OCR do the reading, and you've got text you can translate, hear, and mine for vocabulary — all in one workbook. (Born-digital after all? The PDF guide covers that path.) It's free to start, no credit card; the paid plans simply raise the monthly limits if you're working through a lot of scanned pages, and you can see how the limits work on the pricing page. Grab that page you've been meaning to study and give it a try.