Fast local book reading with free Chrome translation

EPUB, PDF, TXT, Markdown, FB2, RTF, HTML, MOBI, AZW3, and CBZ/CBR/CB7/CBT comics. A reasonable spread, if we say so ourselves.

Also comic archives (CBZ / CBR / CB7 / CBT): a container of page images with no text layer, opened page by page with OCR forced on, so the speech-bubble text is recognized and laid over each page as translatable plates. CBZ and CBT need nothing extra; CBR and CB7 require 7-Zip installed (the same external-tool pattern as MOBI needing Calibre).

Also a standalone image (PNG, JPG, JPEG, WebP, GIF, BMP, TIFF, JPEG XL (.jxl)): the app OCRs it and lays translatable text plates over the picture, so Chrome/Edge page translation works on it in place - the same overlay the browser extension produces. Needs the Tesseract OCR engine (see -ocr-lang).

JPEG XL (.jxl) pictures are decoded locally with the bundled codec into PNG before display and OCR, including pictures inside books and comics. Transparency is preserved; animated JPEG XL uses its first displayed frame. No additional image program or browser codec is needed.

MOBI and AZW3: requires Calibre installed. CBR and CB7 comics: require 7-Zip installed (CBZ and CBT need nothing extra). DRM-protected files are not supported - the converter respects locks it cannot pick.

Plain-text (.txt) input is decoded by sniffing its bytes - a UTF-8/UTF-16 byte-order mark, then valid UTF-8, then a legacy Cyrillic code page (Windows-1251, KOI8-R, CP866) by detection - so a DOS-era or Notepad "Unicode" .txt reads as text, not mojibake. An unreadable binary (a .docx, .djvu, or a comic archive with no 7-Zip) is refused with a named format instead of being turned into a garbage document.

The generated HTML carries a small, fully client-side reader layer - no server to babysit, works happily on file://:

• Reading themes - a toggle in the navbar (and on index.html) cycles Light / Sepia / Dark / Night. Your choice is stored in localStorage and applied on every page.
• Reading position - your scroll is saved per book, so the app remembers where you stopped even when you don't. index.html shows a "Continue reading" link to the last page you were on, and the navbar carries a thin progress bar across the whole book - one that reads as a percentage on hover or keyboard focus.
• Reading comfort - line spacing and column width are selects in the same navbar, remembered with your theme and applied only once you pick one. A night-mode button swaps the light and dark themes in one click, a reset button returns the text size to its default, and the arrow keys and space bar turn pages at the scroll edges, just like PageUp/PageDown. Comics and scanned pages gain a per-session Fit width mode beside the Ctrl+wheel zoom, so a page fills the window without hand-tuning a percentage.
• Search - the navbar's search panel finds a phrase on this page or across the whole book, and Ctrl+F opens it with the field focused (pressed again inside the field, it gives way to the browser's own find bar); multi-page output ships a local text index, so whole-book search works on file://. Results list context lines and the chosen match is highlighted; recognized OCR text over images is searchable where recognition produced it - image-only pages have no text to match.
• Spending guard - for paid engines, -max-cost N turns the cost estimate (chars / 1e6 * $20) into a hard pre-flight limit: if the estimate exceeds N, translation is skipped and the book is still produced untranslated.

Single-page documents keep a continuous reading view and show a compact contents panel when authored navigation, PDF bookmarks, or headings are available.

The app works out of the box for EPUB, TXT, FB2, RTF, HTML, Markdown. The tools below are strictly optional - install them only if you want full PDF image support and the extra formats.

Without pdftotext: built-in Go PDF reader is used (slightly lower quality). Without ffmpeg: JPEG2000 images in PDFs will not be visible in the browser. Without Calibre: MOBI/AZW3 files cannot be opened. Without 7-Zip: CBR/CB7 comics cannot be opened.

All optional. The defaults are sensible, so you can ignore this whole table until you have an opinion.

Every option above is also available in the doc-html-ui graphical app - file picker, drag & drop, TOC depth, spending guard and a default-handler toggle (opt-in, off by default; the app always adds the "Convert to HTML" right-click entry). On the Microsoft Store build the GUI is the launchable entry point, while double-clicking a file still runs the command-line converter.

During conversion, the GUI shows the current stage beside Convert and Cancel. It shows page or image counts when known, without guessing a percentage or remaining time. The log retains details, and the final status distinguishes reuse, warnings, cancellation and failure.

When a conversion goes wrong, a description from memory is rarely enough to reproduce it. The app keeps its recent run logs on disk - up to 20 runs or 20 MB, oldest trimmed first - so you can pack them into one archive for the author, even an hour later or after a crash.

Available in

Desktop appMicrosoft Store appBrowser extension

You need

A mail program, for the pre-addressed message. The Browser extension has no log store and copies a short summary instead (see below).

1. Open the About page of the doc-html-ui app, its Support and diagnostics group, and press Send logs to the author. One press builds a single archive, opens its folder with the file selected, copies its path to the clipboard and opens a pre-addressed message in your own mail program.

   

2. Attach the archive to that message and press Send. The app never sends anything itself.

The archive is written to %LOCALAPPDATA%\doc-html-translate\reports\ and contains exactly three things:

• environment.txt - version, edition, platform, interface language, whether Tesseract was found and which OCR languages are installed, and the default Ollama model, plus a "logs left out" line with their number when the size limit kept some run logs out of the archive. Always English, so the author reads one format.
• settings.json - the settings that shaped the last run.
• logs/ - the recent run logs, newest first.

Everything in it is redacted first: anything shaped like a key, token, password or Google API key becomes <redacted>, and your own folders collapse to %USERPROFILE% and %LOCALAPPDATA%. Document file names are kept on purpose - a log that cannot be matched to a document usually cannot be acted on. Your Google API key is never in the archive; an automated test fails the build if it could be. The archive is capped at 15 MB, dropping the oldest logs first and telling you how many were left out.

Nothing is ever sent automatically. The app opens no network connection for this; it hands you a file and an unsent message, you attach the archive and press Send. Use Open the archive in the same section to look inside first, and Clear stored logs if you would rather keep no run history on disk at all.

On the command line, -report builds the same archive and prints where it landed.

The Browser extension has no log store to archive, so it offers the equivalent it can honestly give: a Copy diagnostics button in the About block of its options page. It copies a short English summary - extension version, browser, interface language, your option set, and the format, page count and last error of the most recent document - to the clipboard, ready to paste into a mail. It carries no document text and no list of your sites.

Related: Main flags (-report) · Privacy

A .fd-sec file is a FileDO secret file. The app asks for its password, decrypts it with the FileDO installed on your PC, and converts the document or picture inside as it would convert the original. You end up with an ordinary HTML book beside your documents.

Available in

Desktop appMicrosoft Store app

You need

FileDO installed on your PC (winget install SerZhyAle.FileDO, the Microsoft Store, or the FileDO setup or portable build), and the file's password. The Browser extension does not open secret files.

1. Open it like any other document: from the command line, by picking or dropping it in the window, from the queue or the recent list, or with the right-click Convert to HTML entry. Nothing in the bytes says what is inside, so the app only suspects a secret file. First, before asking anything, it decides whether the result is already finished (an unchanged file is reopened with no password at all), whether FileDO is installed and whether the file's size fits the format.
2. Type the password in the masked field and press Open - once for a whole queue; a console asks without echo. An empty password opens only a file made without one: that is obfuscation, not secrecy.

   

The app then decrypts the file with the FileDO installed on your PC into one private folder under your temp folder, converts the document or picture inside exactly as it would convert the original, and removes the decrypted copy as soon as it has been read. A copy left by a crash is swept at the next start.

For unattended runs name an environment variable that holds the password: -fdsec-password-env NAME - a variable name, never the password itself.

Refused, each with its own message: a secret file that holds a folder, one that holds another secret file, one that holds a type this app does not convert, a wrong password (the message offers all three readings: wrong password, not a secret file, or altered), a damaged file, and a FileDO too old for the kind of secret file (update it).

Related: Privacy · Main flags (-fdsec-password-env) · Supported formats