The Master Guide / 02 The brief and research · Culture and dialects
Dialect rules
There is no neutral spoken Arabic. Every spoken or sung line is written for one city's ear and signed off by someone who has that ear.
There is no neutral spoken Arabic. Every spoken or sung line is written for one city's ear and signed off by someone who has that ear.
These hold for every spoken or sung line, from the concept on (decided):
Lock the language in order: market, then dialect, then sub-dialect (Saudi Najdi for Riyadh, Hijazi for Jeddah; Egyptian Cairene or Sa'idi), then the register of each character (age, class, formality, humour). Record it in the project file before any line is written.
Documents in English; spoken and sung words in the client's language and dialect, in their own script inside «…», never mixed with English on a line.
Write in the dialect's own spelling, as people speak. Never Modern Standard Arabic for spoken lines unless the brief asks for it. Read every line aloud: if a real person in that city wouldn't say it that way, rewrite it.
Check for leakage: no Egyptian or Levantine words in Saudi lines, and the reverse; no generic "Khaleeji" where the brief needs a specific Saudi register; no flattened politeness or misfiring sarcasm. Distorted register, more than invented words, is the main AI failure on Saudi dialect.
Plain, common words. Where the everyday term is English ("portfolio"), keep the English rather than an obscure Arabic word.
Write for the ear and the clock: short lines, one thought each, timed aloud at the intended pace, with room for pauses and reactions inside the scene.
Sayings and song: real sayings and pop-culture references, verified with a source. For Saudi work, writing in the style of famous singers (their phrasing, melody shape and sayings) is welcome, and a client page may name them as the style. Never name a real singer in a prompt, and never imitate a living singer's voice closely enough to pass as them.
Sign-off, by dialect: a native speaker of that dialect passes every line. Your own ear counts only for your own dialect and for Modern Standard Arabic (MSA). Every line is a draft until it is accepted, and the project file records who accepted it and when, for the written line and later for the voice:
Your own dialect, and MSA: your own ear, recorded as your acceptance ("director acceptance"), never as "native-checked". When Egyptian Cairene is not your own dialect, a native Cairene reviewer checks every Cairene line first, then you accept it;
Sa'idi (decided): a Sa'idi listener named for the project checks every line, then you accept it. Until one is named, a native Egyptian ear decides (yours, if Egyptian is your own dialect) and the line is marked "not native-checked";
Saudi, Gulf and every other dialect: a native reviewer of that exact sub-dialect checks every line, then you approve it. Until then the lines stay drafts and never reach the client.
Two spellings: client pages show the clean approved spelling; the vowel-marked spelling made for the voice engine stays internal (see Voice and Arabic dialects → Spelling for the engine).
The voice: an AI voice first; a professional voice artist of the dialect is the fallback where no shortlisted voice passes the native listener at casting. The reviewer who accepted the written line judges the audio too (see Voice and Arabic dialects → The native reviewer).
Sources
Rani Amayri (Impact BBDO), Campaign Middle East, "The handbook of Arabic copywriting" (no "white Arabic" for spoken work: choose the market, then the dialect).
Campaign Middle East, "How transcreation fixes what translation breaks".
Al-Sumaidaee et al., Saudi dialect benchmark (2026: models fail mostly by distorting register).