# 5mdld/anki-jlpt-decks: a 10,600-word JLPT N1-N5 Anki deck with audio and example sentences > The deck covers all five JLPT levels, ships word and sentence audio, and is published free on GitHub, AnkiWeb and Markji. It is a content pack, not software, so the interesting questions are about import behaviour, update merging and what the deck cannot do. **5mdld/anki-jlpt-decks** — 【egg rolls】JLPT N1～N5 一万词 Anki 牌组【简／繁】【语音＆例句】；Anki，日语，日文，日本語，牌组，卡组，牌組，卡組，能力考，日檢 - Repository: https://github.com/5mdld/anki-jlpt-decks - Website: https://ankiweb.net/shared/info/832276382 - Stars: 3,014 · Forks: 114 - Language: HTML - License: NOASSERTION - Published: 2026-09-24 · Updated: 2026-09-24 · Language: en - Canonical page: https://hysenlabs.com/projects/5mdld-anki-jlpt-decks ## What the deck actually contains, and who it is built for This is a vocabulary deck, not a study application. The README describes roughly 10,600 entries spanning JLPT N1 through N5, each with a definition, at least one example sentence, a word audio file and a sentence audio file. Chinese display can be switched between simplified and traditional. The intended user is someone who already runs Anki and wants a single importable Japanese vocabulary set rather than assembling one from several sources. The ordering is worth reading carefully, because it is where the deck's claims get soft. N4 and N5 are ordered by the author's own judgement of difficulty. N3 and above are sorted by JPDB v2.2 frequency, and then bucketed by how often each word appeared in past JLPT papers from the 2010 format change through December 2025. The bucket thresholds differ per level: high frequency means four or more appearances at N1, five or more at N2, and eight or more at N3. The README states plainly that the ordering and classification are not guaranteed accurate and are for reference only. Treat the frequency tags as a study hint, not a syllabus. ## How the cards are structured and how the audio is produced Each note carries tags that do the work the fields do not. A tag like N3高频 encodes level and importance. Separate tags mark katakana words, onomatopoeia, pitch accent (a circled number such as ②), and part of speech, where [他動1] means a transitive verb of the godan class, 2 marks ichidan, and 3 marks カ変 or サ変. Bracketed markers split the content: [例] for an example sentence, [関] for synonyms or related words, [対] for antonyms, [補] for supplementary notes. Two buttons sit on the card: a search icon that jumps to an external dictionary, and a feedback link to the GitHub issues page. Highlighting is colour-coded inside example sentences. Orange-yellow marks the target word, teal marks a synonym or related word, blue marks an antonym. Inflected words usually highlight only the stem. Word audio is described as mostly recorded by human announcers, while sentence audio was generated with VOICEPEAK and then checked by hand. The card shows two playback buttons: a black-and-white one for the bundled local audio and a purple one for automatically generated online audio. The README explicitly asks users not to report pronunciation problems for the purple button, which tells you the project treats that fallback as out of scope. ## Installing the deck and importing it the first time The README's install path assumes official Anki. It does not promise compatibility with older Anki versions, and it states that AnkiApp and AnkiPro are not supported. Android users are told to avoid the AnkiDroid v2.16.0beta2 build distributed through "Anki 中国" because of compatibility problems. Download the latest .apkg from the Releases page. The README warns to take the .apkg file and nothing else from that page. ```bash # No install command exists. The deck is a file you import. # 1. Install official Anki from https://apps.ankiweb.net/ # 2. Download the newest .apkg from # https://github.com/5mdld/anki-jlpt-decks/releases # 3. In Anki: File > Import, then select the .apkg ``` On the first import the README recommends enabling the option to import all deck preset configurations, so the deck's own scheduling presets come with it. After importing, review the deck options through the gear icon next to the deck name; the README links video walkthroughs for Android, iOS and desktop rather than listing every setting in text. If you want multi-device progress and cloud backup, sign up for AnkiWeb and sync. The README cautions that the first media sync is slow, potentially tens of minutes given the audio volume, and suggests keeping Anki in the foreground and the screen awake on mobile. ## Updating is a re-import, and it overwrites local edits Stable releases arrive roughly every one to two months, and the update method is to download the newest .apkg and import it again. The README says a normal update does not disturb study progress. That is the good half. The bad half is stated just as directly: importing an update overwrites local changes to deck content, tags, fields and templates. For updates, the README suggests setting merge note templates on, update notes to always, and update note templates to always. If you have customised anything, it advises against overwriting and suggests comparing against the repository metadata by hand instead. That is a real cost, not a footnote. Anyone who edits fields or templates is signing up for manual reconciliation on every release. Anki's own mechanics add a second gap. Deck structure changes and deleted cards do not sync automatically with a deck update, so housekeeping is manual. The README gives a worked example: a card sitting in the N3高频 deck but tagged N2高频 can be moved by hand using the tag. It also gives a search string for finding cards that are missing the current version tag. ```text deck:eggrolls-JLPT10k-* -tag:eggrolls-JLPT10k-*::v25.01.01 ``` Results from that search can be deleted or given a del tag to suppress update detection. Version tags on cards carry a date, for example v25.01.01, marking the last update of that card. ## Where this deck is the wrong choice The README is unusually candid that the level assignments are not authoritative. N4 and N5 ordering is subjective. N3 and above are ranked by frequency in past papers, and the project itself says the classification is not guaranteed accurate. If your goal is a defensible claim that a specific word is "N2 vocabulary", this deck cannot give you that, and no frequency-derived deck can. Customisation is the second boundary. If you plan to rewrite definitions, add your own fields, or restructure templates, the documented update path works against you. The README points to a plugin, Special Fields, for protecting custom fields during updates, and to Transfer Scheduling Info Advanced for moving scheduling data. Those are third-party add-ons the project links to, not features of the deck. The third boundary is maintenance expectations. The project announced on 2026-07-30 that it has stopped maintaining its Telegram channel and group for personal reasons, while stating that necessary corrections and updates to the deck will continue. The last push to the repository was on 2026-09-08, and the most recent release is 26.09.06. Corrections are still shipping, but the community support channel described in the repository is gone. ## How it compares with building your own deck or using Markji The project publishes the same vocabulary in two places with different trade-offs. On Anki you get the full card template, the tag system, the local audio, the two-button audio playback and FSRS scheduling. On Markji, an iOS and Android spaced repetition app, the deck is opened directly from its page with no import step. Markji uses an MMX memory algorithm that the README describes as sharing an origin with FSRS, and the project frames Markji as the lighter, faster-to-start option for people who want to begin immediately and keep going. The real difference is control. Anki gives you the fields, the templates and the update merge options, which is exactly why updates can overwrite your work. Markji gives you a managed app experience where that problem does not arise, at the cost of the customisation Anki users expect. The README notes that deck content and presentation differ slightly between platforms because of how each is maintained, so the two are not byte-identical. The other alternative is assembling your own deck from frequency lists and a dictionary export. That gives you full control over ordering and fields, and it removes the re-import problem entirely. It also means sourcing and checking audio and example sentences yourself, which is the bulk of what this project has already done. ## Licence and upgrade cost The repository carries a LICENSE file, but the metadata reports the licence as NOASSERTION, meaning no standard SPDX identifier was detected. The README does not spell out reuse terms in the text available, so if you intend to redistribute the deck, bundle it into another product, or republish the audio, read the LICENSE file in the repository root directly. Nothing here is legal advice. The upgrade cost is predictable: a download and an import every one to two months, plus manual reconciliation if you have edited anything. The README's own update guidance is to back up Anki data before updating or migrating. If you keep the deck unmodified, that is the whole cost. If you customise it, budget time per release, or isolate your edits so a re-import cannot touch them. ## Conclusion Adopt this deck if you want one importable JLPT vocabulary set with example sentences and audio, and you are willing to re-import a new .apkg every one to two months. Do not adopt it if you need guaranteed JLPT alignment, a maintenance commitment beyond bug fixes, or a deck you can freely customise, because the documented update path overwrites local field, tag and template changes. Before importing, check the Releases page for the current .apkg, confirm you are on official Anki or AnkiDroid rather than AnkiApp or AnkiPro, and decide whether you will keep your own edits in a separate note type so the next import does not erase them. ## FAQ ### How do I install the anki-jlpt-decks deck? Install official Anki, download the newest .apkg file from the Releases page, and import it in Anki. The README recommends enabling the option to import all deck preset configurations on the first import so the deck's presets are applied. ### Does anki-jlpt-decks work on AnkiDroid and AnkiMobile? The README says to use the latest official Anki or AnkiDroid, and links AnkiMobile Flashcards for iOS and iPadOS. It states the deck does not guarantee compatibility with older Anki versions or with AnkiApp and AnkiPro. ### Will updating anki-jlpt-decks erase my edits? Yes. The README states that importing an update overwrites local changes to deck content, tags, fields and templates, and suggests comparing against the repository metadata by hand if you need to keep customisations. ### Does anki-jlpt-decks include audio? Yes. Word audio is described as mostly recorded by human announcers, and sentence audio was generated with VOICEPEAK and manually checked. Cards show a black-and-white button for bundled local audio and a purple button for automatically generated online audio. ### Is the anki-jlpt-decks level classification accurate? The README states that the ordering and frequency classification are not guaranteed accurate and are for reference only. N4 and N5 are ordered by the author's subjective difficulty judgement, and N3 and above are sorted by JPDB v2.2 frequency with frequency buckets from past JLPT papers. ## Sources - [5mdld/anki-jlpt-decks on GitHub](https://github.com/5mdld/anki-jlpt-decks) - [Issues](https://github.com/5mdld/anki-jlpt-decks/issues) - [Project website](https://ankiweb.net/shared/info/832276382) - [README](https://github.com/5mdld/anki-jlpt-decks/blob/main/README.md) - [Releases](https://github.com/5mdld/anki-jlpt-decks/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/5mdld-anki-jlpt-decks