01
Record
Speak a phrase for three seconds.
Your phone microphone. No card. No signup.
A phoneme-level coach for learners of English. Speak for three seconds. Read the map.
No adjectives. Every value is a measurement. Read the column, pick a row.
| coverage | 44 / 44 | Phonemes |
| dialects | 23 | en-US · en-GB · … |
| accuracy | 0.96 | Median F1 |
| latency | 540 | Milliseconds |
| format | Web · iOS · Android | Surfaces |
| cost | Free · Beta | No card |
Your phone microphone. No card. No signup.
Three acoustic models vote. You get one number per sound.
Five or six phonemes usually. The rest are already fine.
Phonemes in English.
We score every one.
23 dialects. 1 method.
One number per sound. Not one number per session.
CorePick en-US or en-GB. We measure against that reference, not a mix.
23 DialectsNo confetti. No notifications. Open it when you want.
QuietPhoneticians set the reference. Engineers wired the model.
SourceI had ten interviews this year. Nobody asked where I was from.