JPLearn

A desktop app for Windows · 夜間飛行

Step into
Japanese.

JPLearn's menu is a floating Japanese world — and the learning inside is just as deliberate: spaced repetition, seventeen practice modes, and a tutor that never phones home.

Kana to N1Open source · Apache 2.0Fully offline
The app · 空間

Menus you fly through, not tabs you hunt.

JPLearn's home screen is a 3D Japanese environment. Study areas are places, and moving between kana, kanji and conversation is a camera flight — like the one you're on now.

Spatial 3D menuCamera transitionsYour desktop, not a browser
The route · 道

Six islands, crossed in order.

A 16-node curriculum graph runs from your first hiragana to N1. Blocks unlock at 80% mastery of the one before — and every gate is soft: the app warns, you decide.

あHiragana1 / 6 · ひらがな
カKatakana2 / 6 · カタカナ
漢Kanji3 / 6 · 漢字
語Vocabulary4 / 6 · 語彙
文Grammar5 / 6 · 文法
読Sentences6 / 6 · 例文
Practice · 灯

Seventeen lanterns, one flame.

Every mode is a different light on the same memory — so review never collapses into flipping flashcards. Handwriting is checked stroke by stroke on a canvas; speech is graded by ear.

Recognition
Romaji SprintMeaning MatchCharacter Match
Recall
Stroke OrderHandwritingTyped RecallSpeech Recall
Challenge
Sentence AssemblyParticle ClozeContext ClozeCompound BuilderVibe CheckImposterConjugation Drill
Listening
ListeningDictation
Blended
Interleave Mix

FSRS decides which lantern lights next.

Memory · 引力

Gravity for what you learn.

FSRS models your forgetting curve per card — stability, difficulty, retrievability — and schedules each review for the moment before a word slips out of orbit.

  • 90% target retention, tunable to your pace
  • ×3 priority for due cards, so nothing starves
  • 3 of 5 misses flags a leech for focused drilling
The tutor · 師

A guide who lives here.

The optional AI tutor is a local llama.cpp model sized to your RAM, from 1.4 to 5.5 GB. VOICEVOX speaks Japanese aloud; Whisper grades your speech — all offline, on your own hardware.

On this device
Why does は sound like “wa” here?
As the topic particle, は is read wa — it marks what the sentence is about.
Progress · 星

Your sky fills in.

XP to level 100, streaks that survive timezones, milestone badges, JLPT readiness for every level — stored in a local database, exportable to CSV whenever you like.

  • 8-day streak, UTC-and-local safe (example save)
  • 126 characters and words mastered
  • 68% ready for N5, tracked per level

The gate is open.

JPLearn is built in the open — watching the GitHub repo is how you hear about the Windows download and the Kickstarter the moment they land.

Get notified on GitHub

No mailing list, no tracking — GitHub is the only bell we ring.

The campaign · 物語

The Kickstarter, at a glance.

Straight from the campaign story — real counts from the current build, the plan after funding, and what each tier gets.

What is already in the app, counted from the current build: 10,659 study cards ready on install; 2,218 kanji covering N5 through N1; 8,057 vocabulary words graded by JLPT level; 208 kana; £0 subscription cost as a one-off purchase; 88 grammar patterns with worked examples; 60,000 example sentences searchable offline; and JLPT coverage across all five levels.
From funded to released, in four phases: October 2026 scope lock and review; November 2026 the backer beta opens; December 2026 to January 2027 corrections and release candidate; February 2027 JPLearn 1.0. Dates are targets, not guarantees.
The six reward tiers: One Small Step at £1 gets backer updates; Supporter at £5 adds your name on the Supporter Wall; Early Bird at £15, limited to 150, adds the JPLearn 1.0 Windows licence and beta access; Full Access at £25 includes the same; Founder at £45 adds in-app credits and the Founder feedback channel; Project Sponsor at £100, limited to 30, adds the Sponsor Wall and a private development Q&A. All rewards are digital.

The full story goes live with the campaign — watchers hear first.

When does JPLearn launch?

There's no fixed date yet. Development happens in the open — watching the repository on GitHub is the most current source of truth.

What platforms will it run on?

JPLearn is being built for Windows first. Other platforms depend on interest, so watching the repo genuinely helps that conversation.

Will JPLearn be free?

Pricing hasn't been decided. It'll be shared closer to launch — keeping daily study accessible is a goal, not an afterthought.

Does the AI tutor need the internet?

No. The tutor is optional and runs a locally installed model on your own machine. Explanations work offline, and nothing you type leaves your device.

What happens to my study data?

It stays in a local database on your machine. There's no account, no telemetry, and no tracking — the app works fully offline.

Is JPLearn open source?

Yes — the source, issues, and progress are all public on GitHub under the Apache 2.0 license.