Learn Japanese by speaking it

Most people who arrive at Japanese with an app behind them can read more than they can say. Kana is learned, a few hundred kanji are familiar, the particles make sense on the page. Then a conversation starts and none of it arrives in time.

That gap is wider in Japanese than in most languages, and it is not really about vocabulary. It is that the spoken language runs on things writing does not record. Length, rhythm, pitch, and the small polite ending that decides how the whole sentence lands.

Why speaking practice is different here

Well Versed sends your voice to the model directly rather than transcribing it first. That matters more in Japanese than almost anywhere else, because so much of what goes wrong survives transcription intact.

Say obasan when you meant obaasan and a transcript, more often than not, will tidy it into the word you intended. You have just called someone’s grandmother their aunt and nothing in the app will ever tell you. Drop the beat on the small tsu in kitte and you have said come instead of stamp. Slide the ん into the next syllable and the word is a beat short, which is the sort of thing a listener notices without being able to say why.

A tutor that hears the audio can hear all of it, and can do the thing a teacher does: say the word back to you properly, in the flow of the conversation, without stopping to deliver a lecture about morae.

What about the script

The tutor writes in Japanese script, because reading it is part of learning the language. Underneath, a Latin reading of your own speech is shown by default, so a beginner is never locked out of following along.

That default is deliberate. Someone who can already read kana can switch the readings off in a moment, but someone who cannot has no way to discover the setting exists. So it starts on and gets out of the way when you no longer need it.

Politeness, from the first conversation

Japanese makes you choose a register before you can finish a sentence, which is the part that most often freezes people who know the grammar perfectly well.

Sessions default to teineigo, the masu and desu polite forms, because it is the register you can use with anyone without causing offence while you are still finding your feet. Plain form arrives later, once the masu pattern is solid enough that dropping it is a choice rather than an accident. The tutor holds the same standard in what it says back, so the model in your ear matches the one you are aiming at.

Particles come in by what they do rather than by translation, a few at a time, starting with wa for the topic, no for possession and ka for a question. There is no grammatical gender and no plural agreement to track, which is the one genuine mercy Japanese offers a learner coming from a European language.

Going to Japan

Turn Travel Mode on and the app changes shape for the length of the trip.

Before you go, sessions turn toward rehearsal: the exchanges you are about to have at a konbini counter, at a ticket machine, at the front desk of a ryokan. Then you land, and every morning there is one real thing to go and do, set in the city you are in and pitched at or below your actual level. Order something and pay for it. Ask where something is. Get through one exchange with a stranger.

You rehearse it with your tutor. Then you close the app and go and do it, and afterwards you come back and tell your tutor what happened. That debrief is the part that behaves most like a teacher: you have already done the frightening thing, so the session’s job is to make one real exchange make sense. What they actually said. Whether what you said worked. How to say it next time.

Japan rewards this more than most trips, because the gap between textbook Japanese and what somebody says at speed in a shop is wide, and because a lot of learners arrive with years of study and no experience of being answered at full pace.

There is a syllabus behind it

Conversation practice on its own tends to circle. You get fluent in the handful of things you already say and no closer to the rest of the language, because nothing is deciding what comes next.

Japanese here has a written CEFR curriculum behind it, about 295 items from A1.1 to C1, of which a hundred are capabilities: concrete things you can do, like introducing yourself, asking for something in a shop, or saying what you thought of something. Those are what the app tracks, and the grammar hangs off them as the means rather than the score.

Each one carries a strength that fades if you stop using it. Nothing is marked learned and filed away, which matters more in Japanese than in most languages, because the counters and polite forms you drilled in month two are exactly what has gone by month six. When something starts slipping, the tutor works it back into a conversation rather than stopping to test you on it.

The five Japanese tutors

Every language has five, each from a different place and each sounding like it. You can switch whenever you like, and the app remembers where you are regardless of who you talk to.

• HarutoTokyo

  A standard Tokyo accent, the hyōjungo pitch pattern used in broadcast Japanese. Calm and well paced, and inclined to model natural phrasing rather than dwell on the rule behind it.

• AkariOsaka

  An Osaka Kansai accent, with the lively pitch and warm rhythm of Kansai-ben. Bright and quick, fond of a joke, and the reason a lot of people find practice feels lighter than they expected.

• RenKyoto

  A Kyoto accent, soft and gently rising, unhurried, with an understated elegance. Gives you room to think and treats each phrase with quiet care.

• YuiFukuoka

  A Hakata-ben accent from Fukuoka, the soft friendly lilt of Kyushu with its gentle sentence endings and easygoing musicality. Puts you at ease quickly and talks to you like an old friend.

• MisakiSapporo

  A clear Hokkaido accent from Sapporo, close to standard Japanese but with the crisp, unhurried clarity of the north. Lets a conversation breathe and never rushes you toward an answer.

What English speakers usually find hard

These are drawn from the pronunciation patterns the tutor listens for in Japanese. Because it hears your voice rather than a transcript of it, they are things it can actually notice and work on with you.

Vowel length changes the word

Japanese contrasts one beat against two, and the difference is not decorative. Oba is an aunt and obaa is a grandmother. Toru is to take and tooru is to pass through. Yuki is snow and yuuki is courage. A tutor reading a transcript sees the word you meant. One listening to you hears the one you said.

Pitch accent, not stress

Word identity is carried in a high-low contour rather than in a stressed syllable. In Tokyo speech hashi low-high is chopsticks and hashi high-low is a bridge; ame high-low is rain and ame low-high is candy. Getting it wrong rarely stops you being understood, but it is most of what marks a speaker as foreign, and it is picked up almost entirely by ear.

The rhythm is counted in beats

Every mora takes roughly the same time, so konnichiwa is five even beats, ko-n-ni-chi-wa, not a word with a stressed middle. English speakers import strong-weak stress timing, which is the single most recognisable thing about an English accent in Japanese.

The small tsu is a silence you have to hold

っ doubles the consonant after it and takes a full beat of its own. Kite is come and kitte is a stamp; saka is a slope and sakka is a writer. English speakers tend to skip the hold entirely, which turns one word into another.

ん gets its own beat

The moraic n is not a tail on the syllable before it. Honya, a bookshop, is ho-n-ya in three beats, and sensei is four. English speakers slide the n into the next syllable, which shortens the word by a beat and is immediately audible.

The r is a flap, not an English r

The tongue taps once behind the upper teeth. It is not the curled English r and not a sustained l, and it is close to the d in the American pronunciation of butter. Most learners land on one of the two English sounds and stay there for years, because nothing written down ever tells them they have.