Accent training
Learn any accent.
Measured, not guessed.
Calibrate on seven words in your own voice. Then work a course that scores what you actually produce — against your vowels, not someone else's.
iPhone · in TestFlight · nothing you record ever leaves the device
How it works
Your accent is the measuring stick
Most pronunciation apps compare you to a recording of somebody else. That cannot work: vocal tracts differ in length by twenty per cent or more, so the same vowel comes out at completely different frequencies in two people who are both saying it correctly.
Accentric measures the distance you moved from your own voice. Calibration captures where your vowels actually sit. Every target after that is a position relative to them.
Calibrate
Seven words, said the way you already say them. About a minute. This is the only time the app asks you not to act.
Learn the contrast
Each unit teaches one specific move — dropping an /r/, broadening a vowel — and says plainly where the trap is.
Say it, and see where it landed
Two numbers come back, not one: how far along the move you got, and how far sideways you drifted off it.
The readout
Not a score. A position.
The frame is your own seven vowels, captured at calibration. The dashed line is the move the unit is asking for. The white dot is the recording you just made.
So “seventy per cent of the way across, drifting toward STRUT” is something you can act on. A bare percentage is not.
Courses
Pick the part, not the phonetics
Nobody sets out to learn Received Pronunciation. They set out to play the part it casts them in. The accent is named on every course, so you always know what you are actually learning.
Hear the difference
“Ask the class to dance in the bath after.”
The same sentence in each accent — press play on a course below. Synthesised voices, not native speakers: a model for the shape, not the last word on it.
What it does not do yet
- Vowels, not prosody. Stress, rhythm and intonation are most of what a listener hears as an accent, and none of it is scored yet. The app lists what it is missing rather than hiding it.
- Single words. Connected speech needs forced alignment, which is the next large piece of engineering.
- Consonant rules are taught, not measured. T-flapping and th-fronting get proper lessons, and the app says outright that it cannot hear them.
Privacy
Your voice stays on your phone
Recordings are analysed on the device and discarded. Nothing is uploaded, there is no account, and the app makes no network calls at all. That is not a policy — it is the architecture.