Luha Yini
Chargement de votre session...
Chargement de votre session...
The official alphabet, lessons and a corpus of voices recorded by people from Ngazidja, Ndzuani, Mwali, Maore and the diaspora.
Free. No account needed to learn.
Every recording joins a shared, checked and open corpus, used to learn, to build tools for the language and to train translation and voice models.
Start by learning, record when you feel ready: every voice is checked before it joins the corpus.
A course in short modules: sounds, letters, then words and phrases, with exercises at every step.
Open the courseRead a suggested sentence aloud. Each recording is tied to your dialect.
Enter the studioCertified contributors listen to recordings and confirm their transcription. The quality of the corpus rests on this work.
How quality is ensuredEvery recording carries its dialect. The corpus stays balanced across the islands, and tomorrow's tools will understand every voice.
| Meaning | shiNgazidja | shiNdzuani | shiMwali | shiMaore |
|---|---|---|---|---|
| person | mndru | muntru | not documented yet | mutru |
| banana | ndrovi | ntrovi | trovi | trovi |
| water | madji | maji | not documented yet | maji |
| nose | pua | mpua | pua | pua |
| to go | hwenɗa | hwendra | not documented yet | uendra |
Forms taken from the dictionary. A dash marks a form not documented yet.
What we aim to reach by the end of the pilot phase, with the community of the islands and the diaspora.
No. You can follow the course and contribute at the same time: the course covers the official alphabet and ends with a certification test.
Yes. Your data can be pseudonymised, and any sensitive information is filtered out before publication.
Through three checks: community review, moderation by certified contributors and automatic normalization (implosives, ambiguities, dialects).
You can transcribe existing audio, translate texts, validate other people's contributions, suggest poems or join collection campaigns.
A targeted, themed collection, for example “Tales of Grande Comore” or “Medical vocabulary”. Each campaign has a goal, a dialect, dedicated texts and progress tracking.
That is the goal: datasets and tools will be opened gradually, under free licences.
Ten minutes are enough to record your first sentences.