Right-to-Left Captions in Urdu and Arabic: What Usually Goes Wrong

Urdu and Arabic are written right to left, and both use a connected script where letters change shape depending on their neighbours. Captioning tools that were designed for English and had RTL added afterwards tend to get this approximately right, which is the worst outcome — wrong enough to look careless, subtle enough to ship unnoticed.

The four failures worth checking

1. Letters that do not join

Arabic-script letters have initial, medial, final and isolated forms. Correct rendering requires the font and the layout engine to select the right form for each position. When shaping fails you get a row of disconnected isolated letters — legible with effort, but immediately visible as broken to any reader.

This usually means the font lacks the required tables, or the renderer bypassed shaping. Testing with one word is enough to catch it.

2. Punctuation on the wrong end

In RTL text a full stop belongs at the left end of the line, because that is where the sentence finishes. Systems that append punctuation positionally rather than logically put it on the right, which reads like a sentence that begins with a full stop.

3. Numbers and English inside RTL lines

Digits run left to right even inside right-to-left text, and so do embedded English words. Getting this right requires proper bidirectional handling. Getting it wrong turns "2026" into "6202" or scrambles the order of an embedded brand name — errors that are easy to miss unless you specifically look.

4. Alignment and anchoring

RTL captions should anchor right, so that each new line starts where the reader's eye already is. Centre-anchored captions are acceptable for single lines and become uncomfortable across two or three, because the start position shifts with every line length.

Line length and reading speed

Reading speed research is dominated by Latin scripts, and the numbers do not transfer directly. In practice, keep RTL caption lines slightly shorter than you would in English — the connected script gives fewer visual word boundaries, so long lines are harder to scan at speed.

• Two lines maximum on vertical video.
• Break at grammatical boundaries, never mid-phrase.
• Leave a full line-height of vertical space; connected scripts have taller ascenders and descenders than Latin.