POST /v1/ocrThaana OCR
Printed Dhivehi in photos, scans and screenshots to text, with Loa (0.31% character error on our page exam).
Akuru modelRead Thaana from images, transcribe Dhivehi speech, translate, read text aloud and chat. One key, one credit balance, free credits every month.
Every answer says which model made it. Akuru's own models are marked.
POST /v1/ocrPrinted Dhivehi in photos, scans and screenshots to text, with Loa (0.31% character error on our page exam).
Akuru modelPOST /v1/transcribeDhivehi audio to text with Raalhu P-2.0, trained on Dhivehi speech.
Akuru modelPOST /v1/translateDhivehi ⇄ English with Akuru's own translator, single texts or batches of up to 100.
Akuru modelPOST /v1/read-aloudLoa reads the photo, then a Dhivehi voice speaks it: text and audio in one call.
Akuru modelPOST /v1/speakDhivehi text to speech, with numbers, dates and money read the way people say them.
Partner modelPOST /v1/chatAnswers in Dhivehi with the whole conversation in mind.
Partner modelWith your email: we send a 6-digit code, no password.
Name, contact number and what you will build (organisations add their registration).
One per app. Keep it on your server; revoke it any time.
JSON over HTTPS, examples in curl, Python and JavaScript.
curl https://api.akuru-ai.com/v1/translate \
-H "Authorization: Bearer $AKURU_API_KEY" \
-d '{"texts": ["ހިންމަފުށީ ކައުންސިލުން އީދު
ފާހަގަ ކުރަން ނިންމައިފި"],
"to": "en", "engine": "akuru"}'{
"model": "akuru-bathheli-T-1.2",
"texts": ["Hinmafushi Council decides
to celebrate Eid"]
}For sites on Cloudflare Workers there is a ready translation kit: every page in English under /en/, cached so each paragraph is translated once, and an announcement flow where staff review the English draft before it is published. Ask us for access.