What "sound by sound" actually means
A short tour through how SprechGo turns a recorded sentence into a list of specific consonants and vowels — and why that matters more than a number out of a hundred.
Most pronunciation apps, after you record yourself, give you a number. Seventy-three percent. Or a row of green and yellow dots. Or, increasingly, an AI voice saying great job, keep going.
We have used most of them. We have never been able to act on any of it. Seventy-three percent of what? Which seventy-three? Which sound was the one that dragged the number down? Should you try again? Should you stop?
SprechGo was built around a different idea: the feedback should be small enough to use.
A sentence is a sequence of mouths
When a linguist looks at the word Schule, they do not see five letters. They see three sounds — /ʃ/, /uː/, /lə/ — each one a specific shape the mouth has to make. The tongue lifts. The lips round. The breath goes through the right channel at the right time.
You can describe every word in a language this way. There are about forty of these sounds in German, forty-four in English, around forty-two in Mandarin counted one way. The whole language is made of them.
When SprechGo listens to you say Schule, it does not score the sentence. It splits the recording into those three sounds, compares each one to what a native speaker’s /ʃ/ and /uː/ and /lə/ look like, and tells you which of the three landed and which did not.
What that looks like in practice
You read the sentence. SprechGo plays it back, then highlights one syllable in tarnished gold.
The
/ʃ/in Schule didn’t reach the back of your palate. Your tongue stayed forward — somewhere closer to an English sh in sheep. Try lifting the middle of the tongue toward the roof of the mouth without rounding the lips quite so much.
That is what one piece of feedback looks like. Not good job. Not seventy-three. A specific sound, a specific reason, and something concrete you can do with your mouth the next time you try.
If everything else in the sentence was fine, that is the only thing it tells you. The rest stays quiet.
Why we made it this small
A patient teacher does not give you a list of twelve things to fix in one breath. They give you the one that matters most this week, and they let the rest of your mouth stay where it is. By next week, the one thing has become a habit, and they can show you the next one.
We tried to build SprechGo the way that teacher would talk to you. One thing at a time. The thing that will move you furthest if you change it. Said in language a mouth can hear.
What SprechGo does not do
It will not tell you that you sound natural. Naturalness is a question of prosody, rhythm, and the small unfair things native speakers do that are hard to describe — and harder to teach. SprechGo will get your individual sounds right. The music on top of them you will have to absorb from listening, and from time.
It will not tell you that you sound almost native, either. We do not think almost native is a real category. There is the language you are learning, and there is the version of it your mouth currently makes, and the gap between them is the thing worth measuring.
It will not give you a streak. We do not believe daily reminders are how an adult learns a language. We believe an unhurried hour, once a week, by lamplight, is.
If you spend that hour with SprechGo, what you get back is not a grade. It is a list, sound by sound, of what your mouth actually did — and one quiet suggestion of what to try next.
That is the only promise it makes. It is the one we believed was worth making.