You have a scanned document — a PDF, a photographed page, a paper book you snapped with your phone — and you need it in your language. Capture2Text was built for screen text, but it handles this job well: open the scan on screen, capture each region, let OCR and built-in Google Translate do their work, and assemble the results into a readable translated document. This guide covers the full workflow, the capture tricks that make or break accuracy, and how to turn a pile of captures into a finished file.
You will need Capture2Text v4.6.3 installed — get it on the download page (free, portable, Windows-only). Translation must be enabled first; follow the Google Translate setup guide if you have not. All walkthroughs are indexed in the guides hub.
The workflow at a glance
- Prepare the source. Open the scan or PDF in a viewer at high zoom — Capture2Text OCRs what is on screen, so bigger, sharper text means better results.
- Capture region by region. Press Win+Q, box each paragraph or page section, and let OCR plus translation run.
- Collect the output. Use “Append translation to clipboard” or Capture2Text’s logging feature to accumulate text across many captures.
- Assemble and clean up. Paste everything into a document, fix OCR glitches, and you have a translated version of the scan.
- Optional: make it a real document. If you want a proper searchable PDF rather than a text file, a PDF editor finishes the job (see below).
Step 1 — Prepare the scan for the screen
Capture2Text cannot open files — it only sees your screen. So the scan has to be displayed well:
- Zoom in. In your PDF reader or image viewer, zoom until the text is large and crisp. OCR accuracy climbs steeply with text size; tiny thumbnails produce garbage.
- Maximize contrast. If the scan is gray or faded, many viewers have a contrast/brightness adjustment. Dark text on a clean white background OCRs best.
- Straighten the page. Crooked scans hurt accuracy. Rotate the image upright in the viewer (or a photo editor) before capturing.
- One column at a time. For two-column layouts, zoom so one column fills the screen and capture columns separately — Capture2Text reads left-to-right and will mangle column order otherwise.
- Phone photos of paper books: transfer the photo to your PC and open it full-screen. Photograph pages flat, in good light, with the page filling the frame. This is the same technique language learners use — see Learn Japanese with Capture2Text for capture-quality habits that apply to any language.
Step 2 — Capture and translate, region by region
With the right OCR language active and translation enabled:
- Press Win+Q and drag the capture box tightly around one paragraph or section. Tight boxes beat page-wide boxes — less noise, fewer misreads.
- Finish the capture (Win+Q again, left-click, or Enter). The popup shows the OCR text with the translation alongside.
- Eyeball the original OCR text before trusting the translation. If the OCR misread characters, the translation inherits the error — re-capture that region with a tighter box or higher zoom rather than “fixing” the translation.
- Move to the next region and repeat.
For long documents this is unavoidably repetitive — Capture2Text translates capture-by-capture, with no batch mode. The rhythm gets fast with practice: box, glance, next. Enable Append translation to clipboard (Translate tab) so each paste gives you original plus translation together, which makes the assembly step much easier.
Step 3 — Collect everything automatically with logging
For documents longer than a page, clipboard juggling gets old. Capture2Text has a better option: on the Output tab, enable Logging to save every capture to a file. The log format supports tokens including ${capture} (OCR text), ${translation}, and ${timestamp} — a format like ${capture}${linebreak}---${linebreak}${translation}${linebreak}${linebreak} gives you a running bilingual transcript of the whole document, no pasting required.
Keep the log file open in a text editor on a second monitor (or alt-tab to it) so you can spot-check quality as you go.
Step 4 — Assemble the translated document
- Paste or open your collected text in Word, Google Docs, or any editor.
- Decide your layout: translation only (cleanest for reading) or side-by-side bilingual (best for study or verification).
- Fix the predictable OCR glitches: line breaks in odd places (disable “Keep line breaks” on the Output tab if you prefer flowing text), misread characters in low-quality scans, and headers/footers captured by accident.
- Proofread against the original for anything critical — machine translation of OCR text is a draft, not a certified translation.
Step 5 (optional) — Turn it into a proper searchable PDF
A text file is fine for reading, but sometimes you want the result as a real document — a searchable PDF you can share or archive. That is where a dedicated PDF tool takes over from Capture2Text:
- Adobe Acrobat — the industry standard for creating, editing, and assembling polished PDFs from your translated text.
- Wondershare PDFelement — a more affordable PDF editor that handles creating and editing PDFs well for this kind of assemble-and-export job.
Typical finish: paste your cleaned translation into a Word/Google Docs file, export or print it to PDF with one of the tools above, and you have a searchable translated document. (Links above are affiliate links — they cost you nothing and support this independent site.)
Tips that make or break scan translation
- Match the OCR language to the scan exactly — including Simplified vs Traditional Chinese. The wrong dictionary is the #1 cause of gibberish output.
- Tables and forms need cell-by-cell captures. Capture2Text does not understand table structure; box each cell or row individually.
- Handwriting is out of scope. Tesseract handles printed text. Handwritten scans will produce nonsense — no capture trick fixes that.
- Vertical Japanese text: set the text direction (Win+O toggles horizontal/vertical/auto) before capturing manga or traditional-layout documents.
- Internet is required for the translation step — the OCR step works offline, but translations will silently not appear without a connection.
- Do not capture sensitive documents with translation on. OCR text is sent to Google’s servers for translation. For confidential scans, disable translation and translate sensitive passages separately.
For the general capture technique that underlies all of this, the screenshot-to-text workflow guide is a good companion read.
FAQ
Can Capture2Text translate a PDF directly?
Not by opening the file — Capture2Text only OCRs what is on your screen. Open the PDF in a reader, zoom in, and capture each region. The workflow above turns those captures into a translated document.
How do I translate a whole book?
Page by page, region by region, with logging enabled so captures accumulate into a file automatically. It is manual work — there is no batch mode — but the per-capture rhythm is fast once set up. Photograph or scan pages flat and well-lit for best results.
Why is the translation good for some pages and bad for others?
Almost always OCR quality, not the translation engine. Low contrast, small text, crooked pages, or the wrong OCR language produce bad source text, and both Google Translate and any other engine inherit those errors. Fix the capture, not the translator.
Can I keep the original and translation side by side?
Yes. Enable “Append translation to clipboard” and each paste contains both, or use logging with ${capture} and ${translation} tokens to build a bilingual transcript automatically.
Do I need to pay for Acrobat or PDFelement?
No — they are only needed if you want to produce a polished, searchable PDF at the end. The entire capture → OCR → translate → collect workflow in Capture2Text is free, and a plain text or Word document costs nothing.