Arabic AI voiceover

The Arabic voices, with samples

Each sample reads the same line so you can compare them fairly.

ar-XA means Standard Arabic, deliberately

The mapped locale is the multi-region ar-XA rather than ar-EG or ar-SA, and the trade-off is clear: every voice reads Modern Standard Arabic that any Arabic-speaking market understands, and none of them performs a dialect. For a pan-regional product video that is the right call. If a country-specific voice matters to you, the per-scene voice dialog searches the full Google catalogue, though the director's script will still be written in MSA unless your brief asks for dialect phrasing, which the voices will then read as written.

The same eight characters, reading right to left

All eight premium characters exist in ar-XA: Puck and Laomedeia for the energetic ad read, Charon and Erinome for instruction, with tutorials pinned to that pair mechanically, Umbriel and Sulafat for warmth, Iapetus and Aoede for the neutral register. The director casts one by style and gender from your brief, and any scene can override it. Because a premium voice name embeds its locale, the Arabic cast never leaks into another language's scenes if you build a bilingual video by hand; each scene's clip is generated by its own voice.

What right-to-left changes on screen

The moment the script is written in Arabic, the composition detects the right-to-left text and mirrors itself: authored left alignments become right alignments, and the caption pill lays its words right to left with the first spoken word at the right edge. Numerals and Latin trademarks embedded in an Arabic sentence are the residual risk, since a mixed-direction line can order unexpectedly inside a centred pill; keeping such tokens at the start or end of the sentence, or turning captions off for that scene in the voice dialog, both work. Verify with one render rather than assuming.

Pacing, credits and script hygiene

MSA reads measured and slightly formal, so lines run longer than conversational English; scenes stretch to their clips automatically and a long line finishes over the next cut under dynamic pace. Per scene you control a start delay in seconds and speed steps from 0.85x to 1.25x, and the music bed ducks itself to 22 percent under speech. Narration is billed by length, about 1 credit per 330 characters. Write numbers as words when their spoken form matters, respell foreign names in Arabic letters to steer pronunciation, and know that generated narration never carries an em dash by rule.

Captions in this script

When the script is Arabic the render flips into right-to-left mode: the caption pill sets an RTL text direction, so words sit in correct reading order, and slot text alignment across scenes mirrors left for right. Word-by-word highlighting still tracks the voice, beginning from the rightmost, first-spoken word. What no look provides is an Arabic typeface; none of the catalogue fonts ship Arabic glyphs, so the render host's fallback font draws the letters, connected and correct but not matched to the design. A line mixing Arabic with Latin brand names or digits can still order oddly inside the pill, so render such scenes once and check before shipping.

Questions people ask

Related pages