Prosody
A language app that teaches the part every other app skips — not just the words, but how a sentence is actually said.
Type:
In-house product
Year:
2025
Platforms:
iOS and web
On the App Store:
Capabilities:
Speech AI, pronunciation scoring
Live at:
Inside the product
The work, on screen
Skills
Six parts of a spoken phrase are scored separately, including rhythm, pitch, flow, and tempo.
Result
The score and level sit above the reference and learner pitch curves the phrase was measured on.
The challenge
Learners put in the hours, finish the vocabulary, and still sound foreign the moment they open their mouths. The words are right and the delivery is wrong — the pace, the pauses, the rise and fall, the syllable that carries the weight. It is the first thing a native speaker hears and the last thing any app teaches.
What we built
Prosody is built on a single loop: hear a native phrase, say it back, be scored on how you said it, go again. Meaning is held back on purpose — a phrase arrives as sound alone, and the text, the context and finally the translation open up as the delivery improves. Progress is earned by ear, which is how the language was learned by everyone who speaks it natively.
In production
Sixteen languages, live on iOS and the web, with scoring precise enough to grade a recording the way a tutor would — every attempt measured against native reference audio, syllable by syllable. Not a prototype: a finished product in people's hands.
