One short dialogue a day. You hear a line, the mic opens by itself, and you say it back out loud. Then you see, word by word, where your rhythm went wrong.
Each line lights up word by word as it's spoken, with your language right under it. Slow it down, loop a line, tap a word you don't know, then hit the mic and say it yourself.
No hearts, no gems, no stress. Every step below shows how long it really lasts, and the session ends while you still want more.
Then the lines you found hard come back on their own, days later. As real lines, not flashcards.
Today gives you one dialogue. It picks by your level, a topic you haven't done this week, and the lines you need to review.
One full pass in Listen mode. The text moves with the voices, a translation is one tap away, and nothing stops.
The line plays, the mic opens, you say it back. Word colours in 300 ms, your full score in 3.5 seconds. Three tries a line, five at most, and a low score never stops you.
80%OF THE SESSION
0:20Lines done, your best line, the line to practise, and one thing to fix tomorrow.
+1 DAYThe lines you found hard come back after 1, 3, 7, 14 and 30 days, ready to say again.
You said all seven words, and a native speaker may still say "Sorry?" Prosody is the score for rhythm and stress. It catches this, so we show it first.
Every choice below comes from a research paper. If there is no paper, we don't make the claim.
Sixteen learners shadowed recorded conversations ten minutes a day, four days a week, for eight weeks. Native listeners found their new, unpractised speech easier to understand and more fluent. A session here has the same shape: ten minutes, one short dialogue between two people, your voice over it. Our dialogues are our own, written for SpeakGOAT and voiced by AI.
Saying the same line again stops helping after four or five tries. So each line gives you three tries, five at most, and then you move on.
The review finds clear gains in how easy you are to understand and in rhythm, but no clear result for single consonants and vowels. So the score puts rhythm first, and we never promise to fix your /θ/.
No ads, ever. No password account.
A2 to B2. Roughly: you understand a lot when you listen or read, but you get stuck when you have to speak. Every dialogue has a level, and Today picks from yours.
Sometimes. Your phone turns your speech into text, on the device when it can, otherwise through Apple or Google. We may also send a recording to Microsoft Azure to score your pronunciation. When you talk to Goat, your recording goes through our server to OpenAI to turn it into text. We don't keep it. See our Privacy Policy.
The annual Pro plan has a 7-day free trial, if your App Store or Google Play account can get one. To finish setup, you start the trial, subscribe or restore a purchase. If your subscription ends, you can keep using SpeakGOAT on the free plan, which has daily limits.
You need to be somewhere you're happy to talk. Headphones and a normal speaking voice are enough. No classroom, no partner, no one waiting for you.
Spanish, Hebrew and Brazilian Portuguese, on every line, tap to show.
Listening trains your ear. Shadowing trains your mouth. That gap is why you understand everything and still freeze.
Early access
Ten minutes a day. Four dialogues to start: 45 lines, A2 to B2, US accents, with Spanish, Hebrew and Brazilian Portuguese translations. Tell us where to send the app.
Get early access