Kleoma is built with artificial intelligence, and we want you to know exactly where.
The voices are synthetic
Every voice you hear in the app is generated by text-to-speech software. No recording of a real person is played to you as if it were a speaker. We use commercial speech services (currently Google Cloud Text-to-Speech, and ElevenLabs for a few languages). We choose from the standard voices these services offer. We do not clone or imitate any real person's voice, and the voices are not presented as real people.
The audio is created in advance by us, from our own lesson text, and stored. When you listen, you are playing a stored file. Your data is not sent to these speech services.
Some lesson text is written with AI
We use AI language models to help draft words, example sentences, translations and dialogues, and to help check them. Our process includes automatic checks and review, but it is not perfect.
- Sentences may be correct but sound unnatural, or in rare cases be wrong.
- Pronunciation or intonation of synthetic voices can be off, especially with unusual words, names or numbers.
- Dialogues are fictional. People, places and events in them are invented examples, not facts.
Treat Kleoma as a learning aid. For important uses (exams, legal or medical translation, official documents), please check with a qualified person.
Telling a mistake to us
If you hear a wrong word, a strange sentence or a voice problem, write to support@kleoma.com with the word or lesson name and the language. We correct content and regenerate audio, and your report helps everyone who learns that language.
Why we say this
Being open about synthetic media is the right thing to do, and the law is moving the same way (for example the transparency rules of the EU Artificial Intelligence Act for synthetic audio). [VERIFY: have a lawyer confirm what labelling the EU AI Act requires for our audio and when it applies; this page and the "synthetic" labels in the app are our current approach.]
Version history
| Version | Date | What changed |
|---|---|---|
| 2026-10 | 2026-10-29 | First version (draft). |