Japanese pronunciation checker. Record yourself and AI rates pitch accent, vowel length, and rhythm.
Select the AI model for audio analysis. Different models may have different capabilities.
Record audio directly from your microphone
To check your Japanese pronunciation, record yourself speaking and upload it: the AI scores you out of 10 and evaluates the features that actually distinguish natural Japanese, starting with pitch accent, the system textbooks barely teach. It identifies how you handle the four accent patterns (heiban, atamadaka, nakadaka, odaka), which decide whether hashi means chopsticks or bridge. It then checks vowel length (short versus long vowels distinguish words like ojisan, uncle, from ojiisan, grandfather), devoiced vowels (the nearly silent u and i between voiceless consonants, why desu sounds like dess), and the Japanese r, a tap that sits between English r, l, and d rather than matching any of them. Mora timing rounds it out, since Japanese gives each mora equal length instead of stressing syllables the way English does. Three drills for the week close the report.
Record yourself speaking Japanese (read sentences plus free speech give the best coverage) and upload it. The AI scores overall pronunciation, then audits the Japanese-specific systems. Pitch accent analysis maps your words against the four patterns (heiban flat, atamadaka head-high, nakadaka middle-high, odaka tail-high) and flags words where a wrong contour changes meaning or just sounds off. Vowel length analysis checks whether long vowels actually get two moras of time, the difference between Tokyo pronounced natively and the collapsed English version. Devoicing analysis verifies that u and i thin out between voiceless consonants (suki sounding like ski). The r assessment listens for the alveolar tap rather than an English r or l. Mora timing analysis checks whether each mora gets even length, including the small tsu pause and the n as a full mora. Three weekly drills target your weakest systems.
Record yourself speaking Japanese and the checker scores you out of 10 across the systems that mark natural speech: pitch accent patterns, long versus short vowels, devoiced vowels, the tapped r, and mora timing. Each system gets its own feedback, and three weekly drills target whichever is weakest.
Yes, against the four patterns (heiban, atamadaka, nakadaka, odaka), flagging words where your contour is off. Courses rarely drill pitch accent, yet it is a big part of why grammatically perfect speech can still sound unmistakably foreign. Getting word-level feedback on it from a recording is the main reason this checker exists.
Because length changes the word in Japanese: obasan and obaasan are two different relatives, and compressing a long vowel does not just shade the accent, it swaps the meaning. The analysis checks whether your long vowels are held for a genuine extra beat, something English speakers shorten almost by default.
Japanese gives every mora roughly equal duration, including the small tsu pause and the standalone n, where English stretches stressed syllables and squeezes the rest. Imported English stress rhythm is among the most common findings for learners, and the timing analysis shows exactly where your beats come out uneven.
If it sounds like an English r, yes: the Japanese r is a quick tongue tap, closer to the dd in an American pronunciation of ladder than to r or l. The assessment hears which substitute you are using, and the drills build the tap, which most learners acquire faster than they expect.
Vowel length, devoicing, and timing read reliably from clean audio. Pitch accent feedback is the most demanding part: it needs clear solo speech, and accent varies by region (the reference is Tokyo standard), so treat that section as guidance against the standard rather than absolute right or wrong. Longer, slower clips get better reads.
What bird is this? Upload a call or song and AI identifies the species and decodes the meaning.
How good is my audio quality? Upload a clip for AI grading of clarity, noise, and frequency balance.
What animal is this? Upload a wildlife recording and AI identifies the species and decodes the call.
BPM finder. Upload any song clip and AI detects the tempo, time signature, and any tempo changes.
Song key finder. Upload audio and AI detects the musical key, chord progression, and modulations.
Scale and mode finder. Upload a song and AI identifies the scale, mode, and any modal interchange.