Read offline. Keep a library. Learn a language.

The simplest way - once re/read is installed and set up in Firefox - is to press and hold on the text of any page. The re/read bubble appears with two options: Read in the reader opens that article in re/read, and Offline reading list takes you to the articles and e-books you have already saved there.

The other way is through the browser's menu: tap ⋮ in Firefox, choose Extensions, then re/read - its panel opens with the same two doors (Read this page in the reader, Offline reading list) and with your saved phrases, highlights and settings.

A third way uses Firefox's own shortcuts: open the offline reading list, tap ⋮ and choose Add to shortcuts. A tile appears on Firefox's start page - the page a new tab opens with - and it opens the reading list; a bookmark to the reading list page works the same way. The tile shows a letter instead of the re/read icon, because Firefox does not fetch icons from extension pages.

What does not work is Add to Home screen in the Firefox menu: the shortcut on the phone's home screen either never appears or, when tapped, shows a message that the app is not installed. That is a limitation of Firefox for Android, reported to Mozilla in 2024 and not fixed yet - it concerns the pages of every extension, not only re/read.

The tile and the bookmark contain the extension's internal address: they keep working after an update of re/read, but after an uninstall and a fresh install they have to be added again.

Yes, on a reader with Android on which Firefox can be installed - an Onyx Boox, for example. In the reading view, open the Aa panel and choose the E-ink theme and the Pages layout. The E-ink theme is black text on a white background, only the greys such a screen can show, stronger highlighter colours and no animations. With the Pages layout the text is divided into pages, and a page turn is one refresh of the screen.

In the settings, in the Pages layout section, you choose how pages are turned by touch and the Page turn effect. The value Dark flash (e-ink) darkens the text area for a moment; on an e-ink screen it is barely visible and it helps clear the traces of the previous page.

First check that the tab is not a private one - in Firefox, private browsing is easy to switch on by accident, especially on a phone.

In a private tab Firefox uses a separate, empty database for the re/read pages - the reading list, the reading view, the saved phrases - and deletes it when the private session ends: the reading list is empty, and an article saved or a book imported in such a tab is deleted with the session.

Nothing is lost - the database of your normal tabs is unchanged. Leave private browsing and open re/read from a normal tab: every article and book is there.

In the browser's database, on your device - in the storage the browser gives the extension. You can look at it in the browser's developer tools (Firefox: Storage, Chrome: Application, under IndexedDB); the databases are called reread-articles, reread-vocab, reread-dicts and reread-models.

The settings and the safety copies of your phrases, highlights and library are kept in the same extension storage, in the part called storage.local.

Deleting an article from the library removes everything that belongs to it - text, pictures, highlights with their notes and the reading position - from the database and from the safety copy alike. The Data on this device row in the Data and copies section of the settings shows how much space re/read's data uses, divided into translation models, dictionaries, the reading list, phrases with highlights and backups.

Saved phrases are not deleted with an article. A phrase you save while reading is kept in your vocabulary without a record of which article it came from, because a phrase belongs to your vocabulary, not to the article: it is underlined on every page where it appears, whichever page you found it on. Deleting the article does not delete the phrase, nor the sentence saved with it: the sentence is stored with the phrase, not with the article. Learned - in the bubble that opens when you click its underline, or in its row on the saved phrases page - moves a phrase to the Learned list; Delete on that list removes it for good, and a button under the list deletes all learned phrases. There is no button that removes the whole vocabulary at once; to keep it before uninstalling the extension, export the backup of everything on the reading list, or the vocabulary alone as a TSV file on the saved phrases page.

A translation model works on whole sentences: the meaning of a word is worked out from the words around it. A single selected word has no such context, so its translation can be arbitrary - a rare meaning, a word cut short, or a word that does not exist. Phrases and whole sentences are translated much better.

For single words a dictionary is the better source. In the settings, under Dictionaries, you can download a bilingual dictionary with one click - for the language you read and the language you translate into, English-Polish for example - or add your own StarDict dictionary from files, a monolingual one too, English-English for example. The dictionary stays on your device. From then on the dictionary entry is shown in the bubble under the translation, with every meaning of the word; click the one you meant and it is saved instead of the model's translation.

The sentence and the dictionary entries are visible in the bubble right away. If you prefer a shorter bubble, switch off Show the sentence and dictionary entries right away in the settings - they are then shown only after a press on More.

A translation model translates only from the source language of the pair you chose. When you read a page in your own language - a Polish page with the English → Polish pair, say - and select a word, the model would get Polish text and answer with nonsense. So before the model is asked, re/read checks which language the sentence around the selection is in.

It uses the language detector built into the browser, available to extensions as i18n.detectLanguage. That is Compact Language Detector (CLD): CLD2 in Firefox, CLD3 in Chrome, Brave and the other Chromium-based browsers. Both are compiled into the browser itself and run on your device, with no model to download and no connection; the text never leaves the browser. We have checked this in the source code of both browsers, not only in their documentation. It has nothing to do with the browsers' own page translation (Chrome's "Translate this page", Firefox Translations), which re/read does not use.

At most one sentence is passed to the detector - the one around the selection, cut to 400 characters - and only when you select text with a translation model switched on. Never the page, its address or its title. Its result counts only when it is confident and names the language you translate into or the language the page declares. The selection is then not translated: the bubble shows the entries of a dictionary of that language, or a message that you have no such dictionary yet. There is one exception: when the selected word is in your dictionary of the language you translate from - an English term in a Polish paragraph, for example - the bubble shows its translation and the dictionary entries. A single word with no sentence around it - a heading, a menu item - is too short for the detector, so it is translated as before, unless the word is in your dictionary of another language. Safari has no such detector, so on an iPhone or iPad re/read skips the check.

Is that guaranteed to stay so? This is how both browsers work today (checked in September 2026) and how they document the call. re/read has no say over the browser's code, so we check the implementation again with every release and describe it in the privacy policy. Should a browser ever back the call with a network service, re/read would stop using it; should the detector disappear, re/read would simply behave as it did before.

The Dictionaries section of the settings offers bilingual dictionaries from WikDict, built from Wiktionary, for pairs such as English-Polish or German-English; one click downloads one. Monolingual dictionaries - English-English, say - and dictionaries from other sources you add yourself: download the files, then in the settings open Add a dictionary from files, pick them, choose the languages and press Add dictionary. Adding from files makes no network request.

re/read reads the StarDict format: files with the same name and the extensions .ifo, .idx and .dict (or .dict.dz), sometimes .syn too. Dictionaries come as archives - .zip, .tar.gz, .tar.xz or .tar.zst - so unpack the archive first and pick those files. A .tar.zst archive needs a program that knows zstd, such as 7-Zip, Keka or The Unarchiver.

Sources we know of, where the dictionaries are free to download and use on your own device - each page states the licence of what it offers:

• reader.dict - monolingual dictionaries built from Wiktionary, free of charge for English, Polish, German, French, Spanish, Ukrainian and about twenty other languages; choose the StarDict format. The bilingual dictionaries on that site are paid.
• FreeDict - around 150 bilingual dictionaries under free licences, English-German and German-English among them, as .tar.xz archives.
• Wiktionary StarDict - monolingual and bilingual dictionaries for hundreds of languages, made from Wiktionary and rebuilt regularly, as .tar.zst archives; English-English, Polish-Polish and German-German are there.
• FireDict's list - a short list of free dictionaries as .zip files: GCIDE, the GNU dictionary of English with 108,000 entries, the French Littré and Académie française, the Swedish-English Folkets lexikon.
• WikDict - the same files the settings download, if you would rather pick them by hand.

If you use KOReader, the dictionaries it downloads are StarDict files too, and they can be added to re/read as they are.

Collections of well-known commercial dictionaries that circulate on the internet are not free to redistribute, and we do not link to them.

Yes, from version 0.5.47. In the settings, under Add a dictionary from a link, paste the address of a StarDict dictionary packed as a .zip file - from one of the sources above, for example - choose the languages and press Download. The extension fetches the archive, unpacks it and adds the dictionary, with the same checks as for files you pick yourself; there is nothing to unpack and no files to choose, which matters most on a phone or a tablet.

The extension connects to that address once, when you press the button, and to nowhere else; it sends nothing of yours along, and only https addresses are accepted. If the address redirects to another server, the file is fetched from there, as a download in the browser would be, and the settings page says so. Archives that are not .zip files (.tar.xz, .tar.zst) still have to be unpacked and added from files.

A dictionary of a million entries takes a few minutes to write into the browser's database even on a fast computer, and on a phone or an e-ink reader it takes longer. Two things help.

On some Android devices - Onyx Boox readers among them - the system throttles an app that has no permission to run in the background, and the import crawls: a third of the dictionary after a quarter of an hour. Open the device's settings for the Firefox app and allow it to run in the background; the same dictionary then lands in a few minutes. Once it is in, take that permission away again - re/read does not need it for reading, and an app allowed to run in the background uses more battery.

From version 0.5.7 the import keeps the screen on while it runs, and an import cut short by the browser being closed or killed is not lost: the dictionary's row in the settings says how far it got and offers Continue, which picks it up from there.

In the settings switch on Use without a translation model: the translation bubble, the underlines, the saved phrases page and the models section go away. Do not install a dictionary (or remove the ones you have) - then selecting a word shows no definitions.

Under that switch, turn on Do not show the bubble when selecting text: on ordinary pages nothing will appear, and in the reading view a selection will only highlight the text; a click or the Escape key removes the selection, Ctrl+C copies it. If you do not want read aloud either, switch off Show the read-aloud buttons.

What you then have: the reading view (Alt+Shift+R or the extension's icon), the offline library with articles, EPUB books and Markdown texts, the highlighter with notes and Markdown export, the export of a document as EPUB or Markdown, search in the document and across the library, the remembered reading position, the safety copies and the backup of everything. Nothing is deleted - each of these settings can be changed back at any time.

Highlights are part of the backup of everything: Export on the reading list writes reread-backup.zip with every highlight and its note, from articles and books alike, and with your notes on whole documents. Import on another device adds them to the same articles and books; the articles are in the same backup, and the books too when the Export with books box is ticked - otherwise import a book from its EPUB file first. A highlight is placed where its words are in the text, so it is in the right place also in a book imported again from its file. There is no separate highlights file any more: the .json export and import the Highlights page had until version 0.5.55 are removed, and an older .json file can still be imported on the reading list.

Export as notes (.md) on the Highlights page writes the quotes as a text file for reading and notes; that file can be imported only as a text (a book in the library), not as highlights.

Yes. The settings page has a Custom CSS section. The rules you type there apply inside the translation bubble, on the reader page and in the toolbar popup - never on the pages you read, and never on the settings page itself, so a wrong rule can always be undone: clear the field and press Save.

Rules that would load anything from the network (url(), @import, @font-face) are refused, so the field cannot become a way for data to leave your device.

The names of the elements come from the extension's own stylesheets, published in the repository on GitHub: src/content/tooltip.js for the bubble, src/reader/reader.css for the reader, src/popup/popup.css for the popup and src/assets/page.css for what every page shares. The settings page links to these files at the version you have installed. The names are kept from version to version where possible; when one has to change, the change is listed in the release notes.