Type Darija in Arabizi, get Arabic. Type Arabic, get Arabizi.
Try one of the examples below, or type your own.
How it works
This engine knows Darija. Not just the alphabet; the language itself. It understands how words sound when spoken, how they're built from smaller pieces, and how they change depending on who's speaking and where they're from.
When you type in Arabizi (the Latin letters and numbers Moroccans use in texts and social media), it figures out what you mean and gives you the Arabic script. When you type in Arabic, it gives you the Arabizi. Both directions, same accuracy.
It handles the things that make Darija tricky to write: the numbers people use as letters (like 3 for ع, 7 for ح, 9 for ق), the French words that have become part of everyday conversation, and all the contractions and shortcuts Moroccans naturally use when texting.
What makes Darija tricky
Darija doesn't have one official way of writing. Someone in Casablanca might spell a word differently from someone in Oujda, and both are correct. The same sound can be written three or four different ways in Arabizi, and people mix French and Spanish words into sentences without thinking about it.
Most tools struggle with this because they try to apply rules from Modern Standard Arabic or French. But Darija is its own language with its own patterns. This engine was built specifically for those patterns. Not adapted from something else, not translated from another Arabic dialect. It was researched and built from Darija itself.
No AI needed
Most people assume you need artificial intelligence to convert between scripts. This engine doesn't use any. No neural networks, no transformers, no machine learning of any kind. It is built entirely from an understanding of the language itself.
This was the hardest part of the project, and the part that took the longest. It meant mapping every sound in Darija, studying how words are formed and broken apart, and documenting how spelling varies across regions and generations. The result is a system that truly knows the language, not one that guesses based on patterns it saw in training data.
This is also why it's so fast. You see the result as you type; there's no model to load, no server to think. The engine handles thousands of sentences per second on ordinary hardware, which means it stays scalable and instant no matter how many people use it at the same time. And because the foundation is a deep map of the language itself, it's what all our other tools are built on.