How to Read Aozora Bunko Books: Editions, Ruby, and Text Files

To read an Aozora Bunko book, open its work card and choose いますぐXHTML版で読む—the link to read the XHTML version now. Start there if you want to read in a browser. Download the text ZIP only when you need a local file or a reading app that accepts Aozora text. The official download guide distinguishes those two formats.

The more useful question is often which Japanese text am I looking at? A modern-character edition of a classic is not necessarily the author's first printed edition. A garbled download may need the correct encoding, not OCR. And the brackets in an Aozora text file may contain reading aids rather than unwanted text.

Start from the work card, not a loose download

The work card, or 図書カード, connects the title, author, source edition, and available files. For example, the card for Natsume Sōseki's Kokoro has both an immediate XHTML reading link and a download table further down.

Open the XHTML link first. Adjust your browser's zoom until the main text and small reading aids are comfortable. Read a few paragraphs before deciding you need a different app. If you can read comfortably here, you can leave conversion for another day.

Keep the work card bookmarked as well as the reading page. Aozora's FAQ explains that a work card's address remains a reference point even when a linked file is replaced. For your reading position, also note the section heading and a short phrase from the paragraph where you stopped; a browser bookmark alone does not reliably describe a place within a long page.

If the title or language is wrong, return to source selection before solving display problems. Our guide to finding the right public-domain ebook edition covers that broader decision.

Check the edition before “fixing” its Japanese

Look for 文字遣い種別, the card's character-and-kana classification. Read the label in two parts: 新字 / 旧字 concerns newer or older character forms; 新仮名 / 旧仮名 concerns modern or historical kana spelling. These are separate choices, not a single scale from “bad text” to “good text.” Aozora's orthography guidelines explicitly allow modernizing characters while retaining historical kana.

For ordinary reading, I would start with 新字新仮名 when the work offers it and modern Japanese spelling is what you know. But that label does not promise simplified vocabulary or a rewritten story. For literary research, choose the source edition your question requires; an old-character label alone does not establish that you have a facsimile of the first printing.

Three Sōseki work cards illustrate why the edition details matter. We checked these records on September 10, 2026; all three classify their text as 新字新仮名.

Those dates belong to the listed source editions, not the first appearance of the stories. Kokoro, for example, appeared in the Asahi Shimbun in 1914. Its card separately identifies the 1995 tenth printing used for input. That is a much more precise description of the digital text's source than “the original 1914 book.”

For pleasure reading, recording the source edition once is enough. If you are comparing a quotation against a paper copy, check the 入力に使用 field—the printing used for input—before assuming a wording difference is an error.

If you download the TXT, open it with the listed encoding

In the download table, テキストファイル（ルビあり） means a text file with ruby reading aids. 圧縮 identifies compression; 文字集合／符号化方式 identifies the character set and encoding. For the three files above, the card specifies JIS X 0208 / ShiftJIS. These fields, not your operating system's default, tell you how to interpret the file. See the official explanation of the download columns.

Use this sequence:

1. Download the ZIP linked from the work card and extract it. Keep the original ZIP unchanged.
2. Open the extracted .txt in an editor or reader that lets you choose the encoding. For these three files, select Shift-JIS.
3. Compare a short passage with the official XHTML. If the result is garbled, reopen the unchanged file with the correct encoding before saving any changes.
4. If another app requires UTF-8, make a separate UTF-8 copy only after the source displays correctly. Keep its source card and edition details with it.

We fetched all three official ZIPs again on September 10, 2026. Each passed its ZIP integrity check, and its TXT decoded successfully using strict Shift-JIS. Strict UTF-8 decoding failed for all three. That is an encoding mismatch, not evidence that these downloads need OCR or that the novels are damaged.

The files inside were kokoro.txt, bocchan.txt, and wagahaiwa_nekodearu.txt. Our three-work edition and encoding record (CSV) includes the source links, dates, filenames, and ZIP checksums. It records a file check—not a successful EPUB conversion or a test on your device. No source novels are included in that download.

One distinction saves needless work: if Japanese characters display coherently but look historically unfamiliar, inspect the edition label. If the text displays as nonsensical strings, investigate encoding. Changing the font or retranscribing the whole book should not be your first response to either problem.

Read ruby and notes without throwing away useful text

Ruby is the small text attached to a word or phrase, often to show its reading. In Aozora TXT, it appears inside 《 》. A full-width ｜ can mark where the text receiving that ruby begins. Aozora documents these conventions in its ruby notation reference.

Here are two sentences written for this guide, not quoted from an Aozora novel:

私は図書館《としょかん》で本を借りた。
｜青い屋根《あおいやね》の家が見える。

In the first, としょかん supplies the reading of 図書館: “I borrowed a book at the library.” It is not an extra word to read after the noun. In the second, the bar starts the mixed-character phrase 青い屋根, “blue roof,” whose reading follows in brackets.

Other brackets do a different job. This third original example contains an emphasis instruction:

まだ間に合う［＃「まだ間に合う」に傍点］。

The sentence means “There is still time.” The bracketed part asks for emphasis dots on the preceding phrase; it is not narration. The syntax comes from Aozora's emphasis notation reference. You can download all three original practice sentences with explanations (UTF-8 TXT).

Do not solve distracting brackets by deleting everything that matches a pattern across the book. Some notes describe characters outside the file's standard character set; removing the note can discard the information needed to identify the character. Aozora's external-character reference shows why these are not all disposable formatting instructions.

For reading, return to the official XHTML when the raw notation becomes tiring. It gives you a useful reference for distinguishing the literary text from its representation in the TXT.

Choose a comfortable reading setup before converting the whole book

If browser reading works, your next improvement may simply be larger text and a way to resume at the right paragraph. If frequent word lookups interrupt you, try the dictionary workflow you already use on a short passage first. Neither task requires converting the whole novel.

For an offline reader or an EPUB conversion, ask a more specific question than “Does it open TXT?” You need to know what happens to this file's encoding, ruby, notes, and section breaks. A plain-text import is not evidence that an app understands Aozora notation. Renaming .txt to .epub does not create an EPUB.

Before committing to that setup, compare:

• An opening passage with the official XHTML, to catch unreadable characters early.
• A passage containing ruby and, when present, an emphasis or unusual-character note.
• A middle section and the final section, to look for order or missing-text problems away from the first page.
• Your actual stopping and reopening workflow, so you can resume without searching the whole novel again.

Use the blank reading worksheet (CSV) for those observations. Keep untested items marked not_checked; opening a file is not a full-book proofreading pass. The accompanying download notes and source attribution explain what our three-file check does and does not establish.

We have not converted these three downloads to EPUB or tested them in a particular reader. Choose a tool based on its current support and your own short comparison, not a claim that every Aozora file will behave the same way.

When the remaining obstacle is language

Once the Japanese displays correctly, decide how much language support you actually need. A dictionary may be enough for occasional unfamiliar words. A suitable existing translated edition may be a better reading choice than preparing a new translation. Neither means the source file needs repair.

Before uploading, adapting, or sharing a work, check its specific rights information and your intended use. Aozora's file-use policy distinguishes works with expired copyright from works still protected, including translations whose translators retain rights. Availability on the site is not a blanket declaration that every work is public domain worldwide.

If you do need a whole translated copy, prepare the book for translation first: preserve the source edition, settle the format, and decide how reading aids and notes should be handled. For a permitted, usable EPUB, BookTranslator's EPUB translator is a translation option. That is a separate step from downloading an Aozora ZIP; this guide does not establish automatic handling of its encoding or annotation syntax.

If you want to keep checking the Japanese alongside another language, the bilingual EPUB guide covers layout and review decisions. Otherwise, return to your chosen XHTML page and keep reading. A correct, comfortable source text is already a successful result.