Lea faka-Tonga is spoken by generations, yet largely unseen by modern speech technology. We're building the open, community-owned recordings that teach machines to hear, read, and speak our language — one voice at a time.
Voice assistants, captions, and translation tools are trained on languages with millions of hours of recordings. Tongan has almost none. The result: the technology our children grow up with can't speak, read, or understand the language of their grandparents — and no company has an incentive to fix it. So we're doing it ourselves, in the open.
Mainstream speech tech skips smaller languages. An open corpus is the only way Tongan gets a seat at the table.
Recorded by fluent, native speakers across every island and community — so the dialects and diacritics stay true.
Released under a permanent open licence so anyone — researchers, teachers, app makers — can build on it, with credit.
Hear why this project started and what it means for keeping lea faka-Tonga alive for the next generation. Watch, then lend your voice.
Watch on TikTokEvery consented recording joins a single open dataset. That corpus trains a family of freely-shared models — each one a tool your community can actually use.
Read any Tongan text aloud in a natural voice — for learning apps, accessibility, and preservation.
Turn spoken Tongan into written Tongan — captions, transcripts, and search for a language that had none.
Bridge the gap for a generation that understands more than it can read — translation grounded in real speech.
No account, no app to install. If you speak Tongan, you can help — right from your browser.
Read the plain-language terms and confirm you're happy to share your recording under an open licence.
A Tongan phrase appears. Tap record, say it naturally, and listen back — as many phrases as you like.
Your clip is reviewed, then published to the open dataset — helping train tools for everyone.
The site, tooling, and models are open source — free to use, adapt, and build on commercially.
Every recording is contributed with explicit consent and shared openly, with credit to the community.