Guide
How to subtitle a video in Darija, properly
Most captioning tools are built for English. On Moroccan Darija they return disconnected Arabic letters, reversed lines, or "correct" the dialect into Modern Standard Arabic nobody speaks. Here is how to get captions that hold up, and what changes when a sentence mixes Darija and French.
Why Darija breaks most tools
Three problems recur. Arabic script is cursive: every letter changes shape with its position, and a renderer that ignores this shows isolated letters nobody can read. Direction comes next: a right-to-left line containing a French word must reverse that word, and only that word. Then the dialect itself: a model trained on Modern Standard Arabic hears "wach" and writes "هل", which is no longer what was said.
Add code-switching, ordinary in Morocco: "kanpreparo un truc pour demain". A model that accepts one language per file has to sacrifice one of them.
The method, step by step
1. Start from the original file, not a version compressed for social media: audio quality decides transcription quality far more than the choice of tool.
2. State the spoken language. This is the setting that matters most: a French file sent to an Arabic model comes back truncated or wrong, and the reverse holds too. If the episode mixes both, pick the mixed option rather than either language.
3. Read the transcript before generating captions. Proper nouns, brands and numbers are the most frequent errors, and the most visible on screen.
4. Choose the display: two short lines at most, burned in for social platforms, since most views happen without sound and without anyone enabling captions.
5. Check the result on a phone rather than a desktop screen. Disconnected letters and reversed lines show up there.
Burned-in captions or an SRT file?
For YouTube and a website, an SRT or VTT file is enough and stays editable. For TikTok, Instagram and LinkedIn, burn the captions into the image: those platforms do not always display external files, and a video with no visible text loses its first seconds, the ones that decide the rest.
In Darija, burning in has a second benefit: you control the font. A font with proper Arabic shaping avoids disconnected letters whatever phone the viewer holds.
Arabizi or Arabic script?
Both are defensible. Arabizi ("wach nta mzyan?") reads fast and matches how many Moroccans write online. Arabic script is easier to read on screen for a wider audience, especially outside Morocco.
The rule of thumb: pick according to the account's audience, keep the same choice across a brand's clips, and make sure the tool can produce both from the same audio.
Frequently asked questions
Can a Darija video be subtitled automatically?
Yes, provided the model accepts dialectal Arabic and you tell it the language. A read-through is still needed for proper nouns and numbers.
How do I handle a video mixing Darija and French?
Choose the mixed setting rather than a single language: a model forced into one language translates or drops the other instead of transcribing it.
Why does my Arabic text appear reversed?
The renderer is not applying right-to-left direction, or the font lacks Arabic shaping. The fault is in the editing tool, not the transcript.
How long does subtitling a one-hour episode take?
A few minutes for automatic transcription, then ten to twenty minutes of proofreading depending on audio quality and the number of proper nouns.
Updated on 2026-09-23