Alpha · closed testing
Measured pronunciationLearn to say it right.
Glosso does not just tell you what to say. It listens to how you say it, finds the exact sounds you get wrong, explains why, and shows whether you are improving.
Glosso is in alpha, in closed testing on Google Play. First A1 unit free forever · 1 month free trial · No account required
Not another language app. A coach that measures.
Most apps check whether you picked the right words. Glosso scores how they sounded, down to the individual phoneme, and turns every mistake into the next thing you practice.
Not one number for the whole sentence, but a verdict on each sound you produced.
Glosso says which sound you substituted and where, for example a /θ/ heard as a /t/.
A profile follows each sound over time, so you can watch one move from 43 to 71.
From a microphone to a verdict on every sound.
A speech model trained on many accents turns your recording into a stream of phonemes. Glosso lines that stream up against the reference pronunciation of the exact sentence and scores the differences.
You speak
Read one short phrase. The app records it and, in normal practice, analyzes it on your device.
The model listens
A phoneme recogniser turns the audio into the sequence of sounds it actually heard.
Compared to the reference
What you said is aligned with the expected pronunciation, sound by sound, and the differences are priced in.
/θ/ → /t/Diagnosed per sound
You get a score, the sounds that slipped, and the one correction to try next.
No speech model is infallible. When the audio is too quiet or too noisy to judge, Glosso says it could not evaluate the attempt instead of inventing a low score. See the FAQ for the honest limits.
Four steps, repeated until it sticks.
A short loop that turns an uncertain attempt into a corrected habit.
Speak
Read the phrase aloud. Glosso records it and analyzes the audio on your device.
Measure
Every sound is scored, not just the sentence. You see the words and phonemes that slipped.
Diagnose
Glosso names the problem and explains it in plain language, with the correction to try next.
Train and re-measure
Drills built from your own errors are folded back into the lesson, and your profile tracks whether the sound is improving.
English today. Built to grow.
Glosso starts with English, taught against a British reference. The engine is language-agnostic: a new language is data plus a pack, not a rewrite.
English
A1 to B2 under construction, with a British reference accent and per-phoneme scoring. A1 is the finished part today.
In alphaFrench
Next on the list. The reference phonemizer and scoring rules are already wired.
In preparationMore to come
Other languages follow the same path. Curriculum and pronunciation come first, not a promise list.
PlannedHear what Glosso finds in your voice.
Read one short phrase out loud. Glosso scores it with the same engine the app uses and points at the sounds that slipped, not just the sentence.
A result is not one number for the whole sentence. Glosso scores every sound, names the ones that slipped and tells you what to try next.
- Per-sound verdict. Each phoneme is matched against the reference, so you see exactly where it went wrong.
- Named problem. Not "try again", but "you said /t/ where the word needs /θ/".
- One thing to practice. The next drill comes from your own error, not a fixed list.
The recording demo is in preparation while Glosso is in alpha. It is held back until the scoring endpoint is hardened against abuse, so it is never opened to anyone by accident.
Start free. Go further with Pro.
The first A1 unit is free for good, and every new learner gets a month of Pro.
Starter
Enough to hear the difference for yourself.
- The first A1 unit, free forever
- Pronunciation scoring on every attempt
- Per-phoneme feedback and corrections
- Daily progress and streaks
No account required.
Everything, as it ships
For learners who want measurable, lasting improvement.
- Every level, A1 through B2, as it is released
- Personalized practice built from your own errors
- Phoneme profile with trends over time
- Offline downloads
1 month free, then a paid subscription. Pricing, renewal and cancellation are handled by Google Play.
Quiet, focused, and built around your voice.
We would rather under-promise.
A pronunciation score is a measurement, not a certificate. Glosso tells you what it heard and how sure it is, and never dresses up a guess as a verdict.
- A high score is not the same as a clean one. Our own reference recordings average about 89/100, not 100. A high score means the sounds matched well for that sentence, nothing more.
- Unclear audio is not a zero. If the model cannot hear you well, it says the attempt could not be evaluated instead of showing a low score.
- Your voice stays yours. Practice runs on the device. Cloud scoring, used for a trial, discards the audio as soon as the score is produced.
The things people ask first.
Is my voice uploaded to a server?
In normal practice, no. The speech model runs on your phone and the audio stays there. When cloud scoring is used, for example during a Pro trial, the recording is sent over HTTPS only to produce the score and is discarded as soon as the request finishes. It is never stored, sold or used to train models.
Can the score tell me how good my pronunciation is?
It gives you an honest estimate, not a verdict. The score measures how closely your sounds matched the reference for that sentence and flags real pronunciation problems well, but it is not a certification, not an exam result and not a "sounds like a native" test.
What is free and what needs Pro?
The first A1 unit is free forever, with pronunciation scoring and per-phoneme feedback. Pro adds every level from A1 to B2 as it is released, practice built from your own errors, the phoneme profile and offline downloads. Every new learner gets one month of Pro at no cost.
Which languages are available?
English today, taught against a British reference. French is the next language in preparation, and the engine is built so more can follow as data.
Your voice does not leave your device.
Pronunciation practice runs locally. When cloud scoring is used to mark an attempt, the audio is processed and then discarded. It is never stored, sold or used to train anything.
On-device by default
The speech model runs on your phone. In normal practice, no audio is uploaded anywhere.
Cloud scoring, discarded
If a cloud or trial mode scores your attempt, the audio is processed to produce the score and then discarded immediately.
No accounts, no ads
No mandatory account, no advertising, no third-party trackers. Only anonymous, first-party product events, kept on our own servers.
Glosso is on Google Play.
Glosso is in alpha, in closed testing on Google Play. Join the test, try the first A1 unit for free, and take a month of Pro at no cost.
Get it on Google Play