Skip to content

Releases

Aobana Reibun 1.8

Choose a tag to compare

@Wyzmic Wyzmic released this 02 Nov 13:42

Aobana Reibun 1.8

The fields you pick on the run dialog are kept, across tabs and when you close it, and Aobana gets
your deck list faster.

Changed

  • Fields are saved when the run dialog closes: a Query, Reading or Write to field you pick is
    kept when you close the dialog without Run, as its other choices are, and is there the next time.
    On notes of one note type, it is kept for that type, as after a run. Opening the dialog and
    closing it without picking a field changes none.
  • Aobana gets your deck list faster: when Aobana imports your known words from Anki, the add-on
    lists your decks in one call to Anki instead of one per deck.
  • Longest and Smallest send each search once: when no sentence fits the length limits, a note
    no longer sends Aobana the same search a second time. The sentences found are the same.

Fixed

  • A field picked on a tab was lost when you switched tabs: on the run dialog, a Query, Reading
    or Write to field chosen on one tab went back to the last run's when you went to another tab and
    came back. It now stays. A source under If none found uses the Reading field picked on its own
    tab.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.7 are kept.

Aobana Reibun 1.7

Choose a tag to compare

@Wyzmic Wyzmic released this 01 Nov 11:19

Aobana Reibun 1.7

Append hotkeys for every source, a Settings button when a hotkey finds nothing, and a faster start
and save.

New

  • Append hotkeys: Ctrl+Alt+W (Aobana), Ctrl+Alt+O (Nadeshiko) and Ctrl+Alt+K (Immersion
    Kit) add a sentence under the ones the card has, never one it already has. Its image and audio go
    after the card's. The Ctrl+Shift hotkeys still replace. Settings → General now shows them in two
    boxes, Reviewer hotkeys: Replace and Reviewer hotkeys: Append; a blank one is off. An Append
    hotkey that finds nothing does not try the sources under If none found.
  • Settings from a hotkey that found nothing: the message now has a Settings button left of
    OK. It opens the run dialog on that card, on the hotkey's tab, so you can change an option and
    try again: Run fills the card, Cancel goes back to it.

Changed

  • Anki starts faster: the add-on loads 3 small modules at start instead of 8. The rest load
    the first time you open its menu or press a hotkey.
  • A batch freezes Anki for less time: its notes are saved in one write at the end of the run,
    not one by one. Undo still takes back the whole run in one step.
  • Context lines without a doubled gap: no full-width space is added after a line that already
    ends in 。、!?」』) or a similar mark. A book's lines still join with nothing between them.

Fixed

  • A word with ヵ or a rare kanji lost its reading: for 一ヵ月 or 𠮟る, Aobana got the word without
    its reading, and Nadeshiko and Immersion Kit found no sentence "in that reading". Such words now
    keep their reading on every tab.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.6 are kept.

Aobana Reibun 1.6

Choose a tag to compare

@Wyzmic Wyzmic released this 26 Oct 11:53

Aobana Reibun 1.6

Fields remembered for each note type, and a note whose type lacks a chosen field is left as it was.

New

  • Fields remembered for each note type: run on notes of one note type, and Reibun keeps the
    fields you picked for that type. Select notes of that type again and the run dialog brings them
    back, on every tab; the hotkeys use them too. Your preferences no longer go when you switch
    between note types. With several types selected, a field they all saved alike stays, and the
    rest are your last run's.

Changed

  • Settings shows the Aobana URL, no longer a field to type in: http://127.0.0.1:5010. Reibun
    starts its own Aobana on that port for each run, so your Aobana app keeps its own; the hover says
    so. A port typed into an older version's Settings is no longer used. If another program
    already uses 5010, Reibun moves to the next free port, shows it in Settings and keeps using it.
    Restore Defaults puts 5010 back.

Fixed

  • A note whose type lacks a chosen field could be half filled: a run with Sentence1 on notes
    whose sentence field is Sentence wrote a new audio clip and kept the old sentence, so the card
    played another line, and the summary counted the note as updated. Such a note is now left as it
    was, on every tab, and the summary names the field: "Skipped, the note type
    has no Sentence1 field: 13". In 1.5 only a note with none of the fields was skipped.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.5 are kept.

Aobana Reibun 1.5

Choose a tag to compare

@Wyzmic Wyzmic released this 22 Oct 18:47

Aobana Reibun 1.5

A Reading field Reibun cannot read now skips the note and says why, and several sentences per note
are always different lines.

Changed

  • A Reading field Reibun cannot read skips the note, on every tab, and the summary says
    "Skipped, the Reading field could not be read (write the reading in kana, e.g. たべる)". A hotkey
    says it in a warning. In 1.4 such a note was searched by the word alone, in any reading, with no
    word about it. A word that takes no reading (a kana word like する, several words, a typed
    食べる=たべる, or i+1 only on the Aobana tab) is searched as before.
  • Some shapes 1.4 read are read no more: pitch marks (たꜜべる), half-width kana (マト) and a
    note in round full-width brackets ((的のみ)まと). Reibun now reads what card setups really
    write: kana, katakana, Anki furigana, ruby, Yomitan's pitch fields and a note in square brackets
    ([的のみ]まと). A guess at anything else could take the wrong reading (た・べる would read
    た), so such a note is skipped and the summary asks for the reading in kana.
  • A compound counts only for a word all in kanji: 日本 read にほん still finds 日本語, but 入る
    read いる no longer takes a sentence for 入口 (いりぐち).

New

  • Two more Reading fields understood: a wave dash at either end, as on an affix card (〜てき
    reads てき), and a note after ; in Anki furigana (食[た;h2]べる reads たべる; 1.4 read た and
    found no sentence).

Fixed

  • Several sentences per note could repeat a line: Aobana 2.1 lists a short line once in each
    title it is in, so Shortest with 5 sentences for 了解 gave 了解 from four titles. The sentences
    of a note, and No repeats across notes, are now different lines by their text.
  • A word field on two lines (猫<br>犬, two <div> or <p> lines, a list) was searched as one
    word, 猫犬, on every tab. It is two words now, as Anki shows it.
  • The Reading field's tooltip asks for the reading in kana first.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.4 are kept.

Aobana Reibun 1.4

Choose a tag to compare

@Wyzmic Wyzmic released this 21 Oct 20:46

Aobana Reibun 1.4

The word field read as text on the Aobana tab, more Reading fields understood, and a compound the
word begins counted.

New

  • A compound the word begins counts: with a reading, Nadeshiko and Immersion Kit now keep a
    sentence where the word starts a longer word read whole, as 日本 read にほん in 日本語 (にほんご),
    or 的 read まと in 的場 (まとば). A word at the end of one (的 in 目的) still does not count.
  • More Reading fields understood: half-width katakana (マト), the downstep ꜜ some pitch
    layouts print, and a note in brackets ((的のみ)まと, まと(的)). These gave no reading
    before, and the word was searched alone.

Fixed

  • The Aobana tab searched the word field as Anki stores it: a word followed by a space typed in
    Anki's editor (猫&nbsp;), a word on its own line or a coloured word found few sentences or none.
    It is searched as the word now, as on the other tabs.
  • Readings on separate lines (まと<div>てき</div>, as some editors write them) were joined into
    one, so no sentence matched and the note stayed empty. The first reading is used.
  • Immersion Kit asked twice for a word it has no example of at all, which cost one more slow
    request per note. It asks again only when examples came back and none reads the word that way.
  • Two lines said slightly the wrong thing: the Requests per note for a reading tooltip now
    counts the fallback search's request, and "needs Aobana 2.0" also covers a reading from furigana
    in the word field.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.3 are kept.

Aobana Reibun 1.3

Choose a tag to compare

@Wyzmic Wyzmic released this 21 Oct 01:00

Aobana Reibun 1.3

The word's reading on every source, and furigana in the word field read instead of searched as
text.

New

  • Reading field on every tab: choose the field that holds the word's reading, and every source
    takes only sentences where the word is read that way (的 read まと, not 目的 or 的 read てき).
    Aobana searches by the reading; Nadeshiko and Immersion Kit keep the sentences their own furigana
    reads so. Kana, katakana, Anki furigana (的[まと]), ruby and Yomitan's pitch fields all work. Each
    tab keeps its own choice, a source under If none found uses its own tab's, and the hotkeys read
    it too. Its Settings default moved to General.
  • Furigana in the word field gives the reading too: 的[まと] is 的 read まと. The Reading field
    wins when both have one.
  • No sentence in that reading leaves the note: when no sentence reads the word that way, or the
    reading fits no form of the word (辛い with たべる), the note is left as it is on all three
    sources, and the summary says "No sentence with the word in that reading".
  • Requests per note for a reading (Settings → Nadeshiko): with a reading, Nadeshiko is asked for
    50 sentences at a time until one reads the word that way, at most this many requests (4 by
    default). Most words still cost one request. Immersion Kit asks once more, at 50 examples per
    title, when none of the examples it got reads the word that way.
  • No Sentence language (Settings → Nadeshiko): the Sentence field is always the Japanese line;
    English or Spanish go to the Translation field.
  • The guide is on GitHub: the add-on no longer carries a README, which Anki never shows. The
    full guide is the repository's page.

Fixed

  • A word field in Anki's furigana (的[まと]) was searched with its brackets, so the sentence
    found had nothing to do with the word, or none was found. Brackets are never searched as text now.
  • Nadeshiko's and Immersion Kit's run summary brought Anki's main window up over the Browser.
    Every summary now opens over the run dialog.
  • The summary said "Updated 1 notes." It says "Updated 1 note."
  • The mouse wheel over a Settings drop-down changed its choice while you scrolled the page. It
    changes a drop-down only once you have clicked it.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.2 are kept.

Aobana Reibun 1.2

Choose a tag to compare

@Wyzmic Wyzmic released this 20 Oct 18:16

Aobana Reibun 1.2

For Aobana 2.0: Custom files as a source, a Reading field, i+1 only, and Aobana reading your decks
for the words you know.

New

  • Custom (Aobana tab): your own .txt and .md files, Aobana 2.0's fourth source, beside
    Subtitles, Books and Manga. Ticked by default; Limit to title lists its folders too.
  • Reading field (Aobana tab): choose the field that holds the word's reading, and only sentences
    where the word is read that way are taken (好く read すく, not よく). The first reading counts
    (すく、よく), katakana is read as hiragana, Anki's お 茶[ちゃ] gives おちゃ, and pitch marks and
    dots are dropped (た↓べる, た・べる). A reading that fits no form of the word is searched as the
    word alone, and the summary counts it.
  • i+1 only (Aobana tab, Options): only sentences where the word is the one you don't know yet,
    by the known words imported in Aobana. Exact search is grayed out while it is ticked. When it
    cannot run (no known words in Aobana, an Aobana before 2.0, or Aobana busy), the whole run stops
    and says why, If none found included. A note with sentences but none i+1 is counted apart from
    No results.
  • Known words for Aobana: while Anki is open, Aobana can read your decks for the words you know
    (Aobana's Settings → Known words → Import from Anki). The add-on answers on 127.0.0.1:5020,
    read-only and from this computer only. Settings → Aobana → Known words for Aobana turns it off
    or changes the port; a port already taken is said at start and in Settings.
  • No note numbers (Aobana tab, Options, on by default): a book sentence loses the note number it
    starts with ((27)さやうにて… becomes さやうにて…). A line that is only a number, a divider or
    an answer key is never picked.

Fixed

  • Notes searched while Aobana indexes were counted as errors ("Aobana returned HTTP 503"). They
    are now counted as Not searched, Aobana is indexing your library, or Not searched, Aobana did
    not answer in time
    , and still go on to If none found.
  • The reviewer hotkeys refused a card whose sentence field Settings named, when the last run was
    on another note type. They now take the first field the card has: the last run's, then Settings'.
  • Immersion Kit's sentence field showed (none) on another note type, and its hotkey dropped
    the sentence. The detected field is kept.
  • Web text reached the card as HTML. A tag in a Nadeshiko sentence was run by Anki's reviewer,
    and Immersion Kit's &copy; showed as ©. The text is now written as text; only the highlight is
    kept.
  • Anki markup in a sentence ({{c1::…}}, [sound:…]) made a cloze card or a sound. It stays
    text on the card, also after Anki's editor saves the note.
  • Add context run again on a note that already had its context said Context added: 1. It
    counts only the notes it changed.
  • Immersion Kit sentences follow two more of Aobana's subtitle rules: ◻ is removed and ☏
    becomes ☎.

Aobana versions

The Aobana source is for Aobana 2.0. Aobana 1.9 still gives sentences, but Custom, the Reading
field and i+1 only need 2.0; on 1.9 the Reading field searches the word alone and the summary
says so.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.1 are kept.

Aobana Reibun 1.1

Choose a tag to compare

@Wyzmic Wyzmic released this 02 Oct 12:43

Aobana Reibun 1.1

Fixes for bold, furigana and option combinations, a new Bold ending option, and a summary that accounts for every note.

New

  • Bold ending (Nadeshiko, on by default). With Bold, a verb's or adjective's ending is bold too
    (食べました, not only 食べ), stopping before any particle after it. Untick it to bold the stem only.
  • Notes with nothing to fill are skipped before searching. When every field the run writes is
    already filled, or the note type has none of them, no search is sent. This saves Nadeshiko's quota
    and the time of a search.
  • The summary accounts for every note, with these lines when they apply: Skipped, already filled
    (no search sent)
    , Skipped, the note type has none of the fields, Found, but nothing to write
    and Deleted during the run.

Fixed

  • Nadeshiko with Bold on a conjugated word. On most verbs and adjectives the card got leftover
    markup instead of a bold word, and with Furigana on, the readings were lost. The whole word is
    now bold, its ending included (食べました), and the readings stay.
  • Immersion Kit with Exact search and Bold. The word was never bolded. It is now, in every
    example.
  • A hyphen inside a word is kept. Wi-Fi became Wi――Fi on Nadeshiko and Immersion Kit. A
    hyphen between letters or digits now stays; on Nadeshiko a dash on its own still becomes ――.
  • A note deleted during a run stopped the run before its summary. The run now goes on, and the
    summary counts it.
  • Immersion Kit with Furigana and No (names). A reading could land on the brackets of a
    speaker's name, and the line came out broken. Readings now go on words only.
  • Immersion Kit with No (names). A line whose only match was inside the removed name could be
    picked, leaving a sentence without the word. Such lines are no longer picked.
  • Opening Settings from the run dialog undid the options you had just ticked there. They are now
    kept, and what you change in Settings still applies.
  • Nadeshiko with Append could add a sentence the note already had. Append never repeats a
    sentence now, on all three sources.
  • A hotkey's If none found stopped at Nadeshiko when no API key was set. It now goes on to the
    next source.
  • The missing-key message pointed to config.json. It now says where the key goes: Tools →
    Aobana Reibun → Settings → Nadeshiko
    .
  • Settings uses the run dialog's names for the same options: Furigana, Include context,
    No (names) and Reviewer hotkey.

Install

From AnkiWeb with the add-on's code 1429349152 (Tools → Add-ons → Get Add-ons), or download
AobanaReibun.ankiaddon and open it with Anki, then restart Anki. Settings from 1.0 are kept.

Aobana Reibun 1.0

Choose a tag to compare

@Wyzmic Wyzmic released this 01 Oct 23:29

Aobana Reibun 1.0

The first release of Aobana Reibun (例文, "example sentences"). It started from the code of
AnkiAutoImage by swagercode, with the author's
permission, and is now its own add-on. AnkiAutoImage fills Anki notes with images from web image
search, Gemini image generation and Nadeshiko. Aobana Reibun does one job: it fills notes with
Japanese example sentences, with their audio and images, from three sources. This note
compares it with AnkiAutoImage, item by item.

  • Aobana: your own subtitles, books and manga, searched on your computer by
    Aobana 1.7 or later. It needs no key, has no quota, and
    can fill sentences, their source, furigana, context and a manga page's image.
  • Nadeshiko: the Nadeshiko API, with a sentence, a screenshot and
    an audio clip from anime, live action and YouTube.
  • Immersion Kit: Immersion Kit's sentences from anime, drama
    and games, with a screenshot, audio, the English translation and the title. No key needed.

It has its own name, folder, settings, log, hotkeys and port, so it can be installed beside
AnkiAutoImage: if you want image search as well, install both.

What's out

Everything image-search-shaped is gone. Each item was measured before it was removed.

removed why
Yahoo image search Yahoo Japan answers automated requests with HTTP 403 (bot-blocked)
The Playwright browser fallback Playwright is not part of Anki, so it could never run
Google Custom Search Google ends the API on 2027-01-01 (the original's own notice)
Pexels the module was never called from anywhere in the original
Gemini image generation not wanted for sentence cards
The General, Gemini Image and Legacy Google settings tabs, the provider-order list, Google quota counting they only served the providers above
The image-search and Gemini hotkeys (Ctrl+Shift+G, Ctrl+Shift+U) same
The bundled vendor/ folder (16.96 MB, with Windows-only binaries) Anki itself ships requests, the one library still needed, on every platform

The package is about 63 KB and contains no compiled files, so the same
package works on Windows, macOS and Linux.

What's in

Aobana, a local sentence source (new)

  • Finds your Aobana by itself. Aobana 1.7 or later is found where its installer put it:
    on Windows through the installer's own registration (per-user or all-users, any folder),
    on Linux in ~/.local/share/aobana-app, and on macOS in Aobana.app. It then starts with
    Aobana's own bundled Python. A portable or source copy can be chosen in Settings instead.
    If two installs are found, or the chosen folder is missing or holds an older version, you
    get a message that says so and what to do.
  • Starts and stops Aobana for you. If Aobana is already running, it is used and left
    running. Otherwise the add-on starts it for the run, on its own port (5010), and stops it
    afterwards. It stops only what it started. The console window can be shown, minimized or
    hidden.
  • Sentence, source and furigana. A sentence field, and optionally a source field with the
    episode or book title. Furigana arrives as real <ruby>, which Anki renders. The matched
    word can be bolded, including its conjugated ending but not the particle after it.
  • Subtitles, Books and Manga, any combination, optionally limited to one title (a show, a
    book or a manga series), chosen with Choose… from the titles in your Aobana. Two ticked are
    searched as a set, so the third costs no time (Aobana 1.7 answers that; an older one gets one
    search of all three, filtered here).
  • Manga page images. An optional Image field gets a manga sentence's page image.
  • Context. The lines before and after the sentence (0–10 each) can go into the sentence
    field around it. If filled: Add context adds context to sentences already on your cards.
    It finds each one again in your library and never stacks a second window.
  • Several sentences per note (1–100), all different, from a single search.
  • No repeats: within one run, no sentence goes to two notes.
  • No (names): speaker tags are removed from subtitle lines. A line with two speakers
    becomes 「…」「…」, and a sound cue is kept. Books are never touched, because their
    parentheses are prose.
  • Append never adds a sentence the note already has.
  • Six pick strategies for the main search and for a fallback search with looser limits, and
    a switch to turn the fallback off.
  • Sentence length counts the text only, not furigana readings or most punctuation.
  • A reviewer hotkey, Ctrl+Shift+W, fills the current card.

Immersion Kit, a third source (new)

Sentences from Immersion Kit, whose free API needs no account
and no key. Thanks to Immersion Kit for making it public.

  • Sentence, image, audio, translation and source title, each into a field of your choice
    or none. Image and audio files are named after the sentence (ik_…), so running again
    replaces them rather than piling up copies.
  • Anime, Drama and Games, any combination; Req. Image takes only a sentence with a
    screenshot (a few have none; every one has audio).
  • Furigana (off by default), from Immersion Kit's own readings, as <ruby>. Bold marks the
    matched word. No (names) removes speaker tags, by the same rule as Aobana's.
  • Context, with its audio and images: the lines before and after the sentence (0–10
    each), around it in the sentence field, the same way as Aobana's. Each context line's
    screenshot and audio go into the image and audio fields with the sentence's, in reading
    order, so the card plays the whole exchange. They come from Immersion Kit's media host, so
    they cost no request. Images and Audio turn each kind off, to keep only the sentence's
    own.
  • A pool from one request. One search returns every title's top examples (5–50 per title),
    and the pick happens here, with the same strategies as the other sources, except random
    across all matches.
  • The fallback costs nothing extra, so it ships on. When no sentence fits the length
    limits, it picks again from the same results with looser limits. Unlike Nadeshiko's, it sends
    no second request. Untick it in Settings to leave such notes empty instead.
  • Paced, one note at a time. Immersion Kit allows about one request every 2 seconds, so a
    run takes about 3.5 minutes per 100 notes, twice that with context (one more request per
    note). A progress window shows how far it is and can cancel; what was found is kept. If
    Immersion Kit still refuses with its rate limit, the add-on waits and retries, then stops the
    run with a message instead of skipping every note left.
  • Append never adds a sentence the note already has.
  • A reviewer hotkey, Ctrl+Shift+K, fills the current card.

Nadeshiko, reworked for the current API (2.4.19)

  • Sentences vary again. The API's random order is a seeded shuffle. Without a fresh seed
    per request, the same word returned the same sentence forever.
  • A candidate pool (1–50) and six pick strategies: random from the pool, longest,
    shortest, median, best match, and random across all of Nadeshiko.
  • A fallback search with looser limits when nothing fits. It has its own pool, limits and
    strategy, and a switch to turn it off. Each fallback is a second request, counted against
    the monthly quota, so it ships off.
  • The translation field, as in the original: the sentence's English (or Spanish) translation
    into a field of your choice, following the same Skip / Replace / Append rule as the sentence.
    Settings from the original (nadeshiko_sentence_en_field, nadeshiko_sentence_en_lang) keep
    working.
  • Category filters (anime, live action, YouTube), exact search and bold, all in the run dialog.
  • Furigana (off by default), from Nadeshiko's own word analysis. A sentence without that
    analysis is written as before, without readings.
  • Req. Image & Audio (off by default), like Immersion Kit's Req. Image: only a sentence with
    both a screenshot and audio is picked, in the fallback too. When none has both, the note is
    left as it is and the summary says how many, instead of filling it anyway.
  • Five notes are fetched at a time instead of one after another.
  • The reviewer hotkey, now Ctrl+Shift+O, fills sentence, image and audio on the current card.

Everywhere

  • A run dialog with a tab per source, Aobana first, with If filled: Skip / Replace / Append. It
    remembers each source's last choices, also when closed without running, and it stays open after
    a run.
  • If none found: tick the other sources on a tab to try them in turn, in the order ⇄ sets,
    for the notes the tab's own source found nothing for, or whose search failed. Each writes to the
    fields on its own tab. The run ends with one summary and one undo step.
  • The reviewer hotkeys follow the run dialog: each uses its tab's last fields and options
    (sources, furigana, bold, Exact search, context, Aobana's sentences per note) and the same
    If none found order, which Settings can also set. A source that fails
    passes the card on, and its error is shown with the next one's result.
  • A Settings dialog with four tabs (General, Aobana, Nadeshiko, Immersion Kit), each one page
    in sections. Every setting has a plain label, and the ones that need it have an explanation.
    General also says what each source needs and which fallback costs a request.
  • A search window with Cancel for the reviewer hotkeys. Anki waits for the sentence: the card
    cannot be answered meanwhile, and Anki's own "Processing…" never comes up. If the s...
Read more