Ultimate Manga Reader’s Guide to Capture2Text: Bubble OCR for Japanese Manga

Reading raw Japanese manga (生漫画, “nama” manga) is one of the most fun ways to level up your Japanese — and one of the most frustrating, because you cannot select the text in a speech bubble. Capture2Text, the free open-source (GPLv3) Windows OCR tool by Christopher Brochtrup, solves this with a dedicated Bubble OCR mode: click a speech bubble and the text lands on your clipboard in seconds. This guide walks you through the full setup for Japanese manga, including vertical text, and stays honest about where OCR falls short.

New to the tool itself? Start with our installation and setup guide, then grab the latest build on the download page — Capture2Text v4.6.3 (2022-03-19) is portable, so there is nothing to install.

What is Bubble OCR?

Bubble OCR is a real, documented Capture2Text feature (the official docs describe it as “How to Perform a Bubble OCR Capture”). Instead of pressing the default Win+Q hotkey and dragging a rectangle, you trigger Bubble OCR mode and click directly on a speech bubble. Capture2Text detects the bubble, captures it, and runs Tesseract OCR on it. For manga readers this is a genuine workflow upgrade: bubbles are awkward to box-select by hand, and clicking is far faster across a whole chapter.

Step 1: Install Japanese language data (including vertical text)

Manga text runs top-to-bottom, right-to-left. Tesseract handles this with separate Japanese language data (jpn_vert) in addition to the regular horizontal Japanese data. Capture2Text ships with English, French, German, Japanese, Korean, Russian, and Spanish out of the box; Japanese horizontal data is included by default, but make sure vertical Japanese is available too.

Follow our how to add OCR languages guide for the exact steps, and read Japanese OCR with Capture2Text for a deeper look at what works and what does not with Japanese text. Accuracy depends heavily on the source image: clean digital raws OCR far better than blurry phone photos of a printed page.

Step 2: Learn the Bubble OCR capture flow

The basic loop is:

  1. Open the manga page on your screen (a browser reader, a viewer app, or a scanned PDF).
  2. Trigger Bubble OCR mode (assign it a comfortable hotkey in hotkey settings — the default Win+Q does a manual box capture; set a second hotkey for bubble mode).
  3. Click a speech bubble. The captured text goes to your clipboard.
  4. Paste it into a dictionary, or send it through translation (see next step).

If a bubble is not detected cleanly, fall back to the manual box capture with Win+Q. Narration boxes and sound-effect text usually need the manual box anyway, since Bubble OCR targets round/oval bubbles.

Step 3: Turn captured text into English

Capture2Text has a built-in translate feature powered by Google Translate — read the full walkthrough in how the translate feature works. Two honest caveats: translation requires an internet connection (everything else in Capture2Text works fully offline), and machine translation of manga dialogue — full of slang, dropped subjects, and casual speech — is rough. Treat it as a “what is roughly happening” aid, not a substitute for learning the language.

Honest limitations: where manga OCR struggles

Read this before you trust every capture.

  • Handwritten and decorative fonts fail often. Shonen-style screamed dialogue, stylized sound effects (擬音語), and hand-lettered text have low recognition rates — Tesseract was trained on print-like type.
  • Tiny furigana (reading aids above kanji) are usually too small to OCR reliably; expect them to come out garbled or be skipped.
  • Vertical text needs the right language data. Without jpn_vert, vertical bubbles will misread badly.
  • Image quality is the ceiling. Low-resolution raws, watermarks, and heavy JPEG compression all drag accuracy down. Nothing in settings fixes a bad source image.
  • Capture2Text is Windows-only and portable; it cannot read DRM-locked apps that block screen capture.

Tips for better manga OCR accuracy

  • Zoom in. Reading at 150–200% in your viewer gives Tesseract bigger, cleaner glyphs.
  • Capture at native resolution. Avoid capturing a scaled-down thumbnail; open the page at full size first.
  • One bubble at a time. Bubble OCR is designed for this — resist the urge to capture a whole page at once.
  • Proofread before translating. A single misread kanji can flip a machine translation. A two-second glance at the clipboard text saves confusion.
  • Keep a Japanese IME handy for quickly fixing misread characters before lookup.

Where to go next

Frequently asked questions

Is Capture2Text good for reading raw manga?

It is one of the best free options on Windows for this exact job: the Bubble OCR mode lets you click speech bubbles instead of dragging boxes, Japanese (including vertical text) is supported, and it is portable and free. Handwritten-style fonts and tiny furigana remain weak spots.

Does Bubble OCR work on vertical Japanese text?

Yes, as long as the vertical Japanese language data (jpn_vert) is installed. Without it, vertical bubbles misread badly. See our guide on adding OCR languages.

Can Capture2Text translate manga to English automatically?

It can send captured text through its built-in Google Translate feature, but you need an internet connection, and machine translation of casual manga dialogue is rough. It works as a rough-meaning aid, not a real translation.

Is Capture2Text free and safe for manga reading?

Yes — it is free and open source (GPLv3) by Christopher Brochtrup, portable with no installer. Read our safety breakdown for details on what it does and does not do.

Why does Bubble OCR fail on some bubbles?

Common causes: decorative or hand-drawn lettering, very small text, low-resolution source images, or bubbles with heavy background art. Fall back to a manual Win+Q box capture in those cases.

Does it work on Mac or Linux?

No. Capture2Text is Windows-only. Mac and Linux users should read our honest platform pages for alternatives.