話す · hanasu
You can finish a whole course without ever saying a sentence out loud. Not here: every phrase is heard, repeated into the microphone, and the browser tells you whether you got there or came close.
Hear the phrase, say it, and have something tell you whether you said it right.
Japanese pronunciation is among the easiest an English speaker can come across. Five pure vowels, much like those of Spanish or Italian, with no gliding diphthongs, no long and short vowels that change quality the way «ship» and «sheep» do, and no sounds that force you to shape your mouth in a new way.
What is different is the rhythm. Japanese is measured in equal beats: every syllable lasts the same, including ん and the double consonant, which count as a full beat even though they do not sound like a syllable of their own. That is why «Nihon» has three beats and not two.
And there is an accent, but it is one of pitch, not stress: the word rises or falls. You do not need to master it to be understood, and it is worth hearing it from the start so you do not lock in a flat accent that is hard to shake off later.
None of this is learned by reading. It is learned by hearing the phrase and saying it, which is exactly the cycle this course is built around.
What is worth knowing before you open your mouth for the first time.
a, i, u, e, o, pure and very close to the vowels of Spanish or Italian. The only one that stands apart is u, which is said without rounding the lips. They do not change sound depending on the word, the way English vowels do.
Each syllable takes up one beat, and long vowels take up two. Cutting one short changes the word: there are pairs that differ only in that.
The small っ makes no sound: it cuts and doubles the next consonant. That silence takes up its beat, and skipping it is what gives a beginner away most.
No syllable is said louder: the voice rises or falls. It varies between regions, so there is no need to obsess over it, but it is worth imitating what you hear.
Four steps, and the order matters: first you hear it, then you say it, and only at the end do you write it.
The phrase plays with the device's Japanese voice, and there is a second, slower speed for long ones. In the listening quiz you hear it without seeing the text and choose the meaning.
You tap the microphone and repeat it. The browser's speech recognition transcribes what you said and compares it with the phrase: you get «perfect», «almost» or «try again».
And at the end you type it, in romaji or in Japanese if you have the keyboard for it. Writing it is what proves you have the whole phrase and have not just half-recognised it.
When you finish the activity, its phrases go into spaced repetition and reappear days later so they do not slip away.
Speech recognition is provided by the browser, not by Hanasuka, and that means the audio is sent to its maker's servers (Google in Chrome, Apple in Safari) to be transcribed. It is said here and in the privacy policy because what nobody expects is exactly what needs saying. If you would rather not use it, the rest of the course works just the same.
Japanese is not measured in syllables, as English is, but in moras: units of time that all last the same. Each ordinary kana is one mora, and so are ん, the small っ and the second half of a long vowel. That is why 日本 (にほん) has three beats and 東京 (とうきょう), four, although in English we would count two syllables in each. Combined syllables like きょ take up just one, because the small ょ does not count separately. A good exercise at the start is to clap once per mora.
Respecting that rhythm is not a detail of accent: it changes words. ビル is «building» and ビール, «beer»; おばさん is «aunt» and おばあさん, «grandmother»; きて is «come» and きって, «stamp». English speakers find it hardest to hold the long vowel and the pause of the っ, because in English length alone does not tell words apart and the ear lets it slip by. The solution is to exaggerate them when practising, until the right length comes out without thinking, and to listen to pairs like these until you hear the difference.
In standard Japanese, i and u become voiceless when they fall between voiceless consonants, or at the end of a sentence after one of them: they still take up their beat, but they can hardly be heard. です sounds like «des» and ます, «mas»; すき (liked) sounds like «ski» and した (below), «shta». Nobody writes it that way, but that is how it is pronounced, and anyone who says «de-su» with the full u sounds like a textbook. It is one of the things that brings pronunciation closest to a native speaker's.
It is best not to overdo it, because it is not a fixed rule: it varies with region, speed and speaker, and in slow or emphatic speech the vowel comes back. The way to learn it is not to memorise where each vowel drops out, but to imitate what you hear. That is why it is worth listening to phrases before reading them in romaji, which is exactly what tempts you to pronounce every letter, and paying attention to the rhythm at normal speed more than at slow speed.
Japanese does not mark a syllable with more force, as English does, but with the height of the voice: each word has a pattern of rises and falls. In Tokyo Japanese, 箸 (はし, chopsticks) starts high and drops, while 橋 (はし, bridge) starts low and rises, and the difference also shows in the particle that follows. 雨 (あめ, rain) drops after the first mora. There are pairs that can only be told apart this way, although in practice context almost always clears up the doubt.
For beginners, the advice is neither to obsess over it nor to ignore it. You do not need to study the pattern of every word to be understood, but saying everything in a flat tone, or with the stress accent of English, is what most gives away a foreign accent. What works is imitating whole phrases, with their melody, rather than single words. And it is worth knowing that the app's voice is the device's: its intonation is usually good, but not always perfect.
Speech recognition turns what you say into text, and the app compares it with the phrase. For it to work well: a quiet place, the microphone close by and the phrase said whole, at normal speed and without separating the syllables, because the recogniser is made for natural speech and not for syllable-by-syllable dictation. If it answers «try again» several times in a row, it is worth hearing the phrase again before repeating it. And the browser needs permission to use the microphone, which is usually given just once.
It is worth knowing what it measures and what it does not. The recogniser looks for the most likely words for what it heard, and sometimes corrects an imperfect pronunciation on its own, so «perfect» means you were understood, not that your accent is flawless. Conversely, a noise or an uncommon word can produce an unfair «almost». It is a tool for losing the fear of saying the phrase and checking that it is understood, not a pronunciation exam. And the recognition is done by the browser, not by Hanasuka.
Three exercises per lesson, in the order of the cycle above.
Phrases from the lesson, repeated out loud and compared with what was understood. «Almost» counts neither as right nor as wrong: it is neither of the two.
The phrase plays without being shown and you have to choose —or type— what it means. It is the closest thing to really understanding someone.
You type the phrase in romaji or in Japanese. Whatever shows you know it is accepted: the phrase with kanji, in kana or in romaji.
It is one of the easier languages to pronounce for an English speaker: Japanese has only five vowels, pure and stable, and no sounds that demand a new position of the mouth. What needs work is the rhythm, because every syllable lasts the same, and keeping the vowels pure instead of letting them glide the English way.
By repeating phrases out loud and checking what you said. The browser's speech recognition makes that possible: you say the phrase, it is transcribed and compared with the original. It is no substitute for a conversation, but it breaks the barrier of saying it for the first time.
You need a browser with speech recognition —Chrome and Safari have it— and to give it permission to use the microphone. There is nothing to install. If you do not want to use it, the rest of the course works just the same: listening, writing and the quizzes do not need a microphone.
To the maker of your browser, which is who does the transcription: Google in Chrome, Apple in Safari. Hanasuka does not store the audio or send it anywhere else, and only receives the text the browser sends back. It is declared in the privacy policy.
Like the single r of Spanish in «pero» or «cara»: a quick tap of the tongue just behind the teeth, close to the tt of «butter» in American English. It is never rolled, not even at the start of a word, nor is it the English r, and some speakers bring it close to a soft l. For an English speaker it is one of the sounds that takes the most practice.
Only a few: つ, which joins t and s like the end of «cats», but at the start of a syllable; ふ, which is blown with the lips almost closed and not with the teeth; the r, a quick tap that is neither the English r nor a rolled one; and u, which is said without rounding the lips. The rest has a close equivalent, and what really needs work is the length of the sounds, more than the sounds themselves.
Yes, and very marked ones. Tokyo's is the basis of the standard Japanese that is taught and heard on the news, but that of Osaka and the Kansai region has different intonation, different endings and a different pitch accent. All Japanese people understand the standard, so it is the one to learn first.
Because what the speech recognition understood is very close to the phrase but does not match it completely: a word swapped, a different ending or a lost syllable. It counts neither as right nor as wrong. It can happen from swallowing the っ or shortening a long vowel, so it is worth hearing the phrase again and repeating it calmly.
Five hundred phrases with their reading, their audio and the «when it is used» of each one.
Japanese in Latin letters: useful in the first week, a ceiling from the second.
The verb at the end and the particles in charge. The full syllabus, unit by unit.
You sign in with Google, with Apple or with an email and a password. No card is asked for, because there is nothing to charge.
Start now
