A text-to-speech engine trained on real Arabic — Fusha, dialects, and names — with seamless transitions between Arabic and English within a single sentence, no break in cadence.
Text-to-speechArabic + English, seamlessAuthentic pronunciationData we produced ourselves
How it works
Four pillars. One method.
01
Seamless language switching
The voice doesn't trip when an English word enters an Arabic sentence. The transition is as natural as a human breath.
02
An Arabic-tuned model
An open base model, tuned on Arabic we produced ourselves — not synthetic translations.
03
Name pronunciation
Omani place names, personal names, technical terms — pronounced as their people pronounce them, not as a model would guess.
04
Our own data
Data is the secret. An independent corpus collecting real Arabic, with permission and ethics.
By the numbers
The numbers, without exaggeration.
Languages
Arabic + English
Seamless within a sentence
Model
Open-source base
Specifically tuned for Arabic
Dataset
ArabicWorldDS
Independent corpus
Status
Sample model
Ready for testing
A voice without people's language isn't voice — it's noise.