Subtitling and translating video for multilingual audiences

Many organisations in East Africa make films in one language for audiences who speak several. A testimony recorded in Kinyarwanda may need to reach a donor in Brussels, a ministry audience in English and a regional audience in Swahili. Video subtitles are usually the quickest and most faithful way to do that, because viewers still hear the speaker's own voice.

Subtitles are not a niche feature. Research by BBC R&D and university partners found that around 10% of the UK television audience use subtitles on any given day, and around 6% use them for most of their viewing.¹ YouTube's own help pages make the same point more broadly: subtitles and captions let you share your videos with a larger audience, including people who are hard of hearing and people who speak another language.²

The two words are often used loosely, and in some regions captions are simply called subtitles. The W3C uses captions for same-language text and subtitles for translations.³ Some regions draw the same line by calling them intralingual and interlingual subtitles.³

Captions are written for people who are Deaf or hard of hearing. They are a text version of the speech and the non-speech audio needed to understand the content.³ Translated subtitles are not directly an accessibility accommodation, but they can carry the same information, including non-speech sounds.³ In practice, the best approach is to produce accurate captions first, then translate them, so every language version keeps speaker names and meaningful sounds.

Translating video starts with knowing exactly what was said. Automatic captions save time, but YouTube warns that their quality varies and that they can misrepresent speech because of mispronunciations, accents, dialects or background noise.⁴ It encourages creators to add professional captions first and to review and correct automatic ones.⁴ The W3C is firmer: automatically generated captions do not meet user needs or accessibility requirements unless they are fully accurate, although they can be a starting point.³

Automatic captions are also produced only in the video's default language.² For a film that mixes Kinyarwanda and English, or uses local place names and programme terms, expect to correct a great deal by hand. In our own practice, a native speaker checks the original-language transcript against the picture before any translation begins.

Subtitles compete with the picture for the viewer's attention, so they need to be short and easy to scan. The W3C offers a compact set of subtitling guidelines that work in any language:⁵

A translation does not have to match the original word for word. Rather than squeezing in every word, condense to the meaning the speaker intended, and check the result with a native speaker who has watched the scene.

• Use one or two lines, generally under 32 characters each.
• Start each new sentence on a new line.
• If a sentence must be split, break it at a logical phrase.
• Identify speakers where relevant, using the full name first and a single name after that.
• Put off-screen speech in italics.
• Show meaningful non-speech sounds in parentheses, in lowercase italics.
• Include background music only when it matters to understanding the film.

Timing is easiest to get right once. YouTube can automatically sync a transcript to a video, but only if the transcript is in the same language that is spoken in the video.² A reliable order is therefore to time the checked original-language captions first, then translate inside that timed file so each translated subtitle appears with the right line of speech.

Give translators the film, not only the text. Tone, gesture and what is shown on screen all shape the right wording. Keep a short glossary of names, titles, places and programme terms so that every language version uses them consistently, and ask a second native speaker to review the translation against the picture, paying particular attention to numbers, dates and quotations.

Most captions and subtitles are closed, which means viewers can switch them on or off, while open captions are always displayed.³ The most common caption format on the web is WebVTT, and SRT and TTML are also widely used.³ YouTube accepts many formats and suggests that beginners use basic types such as SubRip (.srt), which carry timing and text only, while advanced formats such as WebVTT and TTML give more control over styling and position.⁶

Closed subtitle files are easy to correct and let one video carry many languages. Burned-in subtitles suit short social clips that play without sound, but they cannot be switched off or changed. Always keep a clean master without text on screen, so new language versions can be made later without re-editing the film.

Subtitles are only part of reaching a multilingual audience. YouTube lets creators translate video titles and descriptions, and viewers can then find the video by searching in their own language.⁷ It also lets you upload audio tracks in different languages to a single video, so a dubbed or voiced-over version can sit alongside the original rather than as a separate upload.⁷

Subtitles assume viewers can read comfortably at speed. For community screenings, radio-style use or audiences with lower literacy, a voice-over in the local language may serve people better. The right choice depends on who you most need to reach.

• Original-language transcript checked by a native speaker
• Captions timed in the original language before translation
• Glossary of names, places and programme terms shared with translators
• One or two short lines per subtitle, broken at natural phrases
• Speaker names and meaningful sounds included
• Second native speaker review against the picture
• Separate subtitle files kept, plus a clean master without text
• Titles and descriptions translated for each language