Common questions

Yes. Start practising in your browser with no account and no card. Your free credits are a one-off pool, not a daily refill. When they run out, register an email and your balance goes to 600 credits. After that, invite a friend to learn with you: every friend who signs up earns you 400 more, with no limit.

Your audio is streamed to Microsoft Azure Speech to compute the score, then discarded. Tono does not keep recordings and does not use them for training. Full detail on the Privacy page.

No. Open the app and speak. Your practice history lives on the device you used, which also means it does not follow you to another browser or phone.

Azure Speech pronunciation assessment scores every word and every phoneme against a native model — not just the sentence as a whole. For English (US) you also get fluency, completeness and prosody.

33. Each one is scored with its own native locale model rather than an English approximation — that is why accuracy holds up outside English. The Languages page lists what each one reports.

Three causes account for most of it: background noise, speaking faster than you would to a person, and a microphone that clips. The word colours tell you which words the engine doubted — if a whole sentence comes back low, suspect the recording, not your mouth.

The International Phonetic Alphabet gives every speech sound one symbol, so /θ/ always means the sound in "think" no matter which language you are reading about. Naming the sound you missed is what makes the feedback actionable — "say it again" is not.

Yes. Type or paste any sentence into the practice page. For longer material, import it as a course and it splits into sentences you can work through line by line.

Yes. AI chat holds a conversation in the language you are learning. Tap any reply to say it out loud and have it scored, or tap a word you do not know to look it up.

Yes, in the mobile browser. The page needs a secure connection and your permission to use the microphone — both are things the browser asks for, not something you configure.

Check that the browser itself has microphone permission for this site, and that no other app is holding the microphone. If the page still cannot record, it will say so and let you listen to the native reference instead of leaving you stuck.

No. The score comes from acoustic assessment — the engine compares the sound you produced against a native model for that language. A language model never decides whether your pronunciation was good. That is why the same sentence said the same way scores the same twice.