All articles

Bilingual Captions: Put Your Clips in Two Languages

A single clip card connected by two plain lines to two identical cards below, illustrating one clip published in two languages.

Bilingual captions are subtitles that carry your clip's words in two languages at once — either both on screen together, or the spoken one swapped for a translation. For a clipper, they are the cheapest way to widen the audience of a clip you have already cut: you translate text, not a video. This guide covers when it pays off, the two formats, the four-step method, and what machine translation still gets wrong on a thirty-second cut.

Why one language caps a clip's reach

The audience for a clip is not everyone on the platform — it is everyone who understands the spoken language. Captions change that, because a short is watched with the sound on but understood without it: a viewer who cannot follow the dialogue can still read one line and get the joke.

76%of online shoppers prefer information in their own language, and 40% will never buy from a site that only speaks another one (CSA Research, 'Can't Read, Won't Buy', 8,709 consumers in 29 countries).

That figure comes from CSA Research's survey of 8,709 consumers in 29 countries and is about buying, not watching, but the reflex is the same and it is stronger for video: when the text is in your language you stay, when it is not you scroll. A clip cut only in its original language deliberately halves its own pool — the viewers who do not speak it never see the hook, and the ones who do never see why it was worth translating.

The point is not to translate everything. It is that the cost of a second-language version is text, not a new edit, so the marginal audience is cheap.

Why dubbing the video is the wrong tool

The reflex when you want a clip in another language is to translate the whole video — dub the audio, or let an auto-translate feature re-voice it. On a thirty-second short, that backfires three times.

First, the voice is the hook. Dubbing replaces the tone, the timing and the emotion of the original with a synthetic read, and the first two seconds are exactly where a short wins or loses.

Second, the timing stretches. A sentence that takes two seconds in English can take three in French or Spanish; the dubbed track slides out of sync with the cut, and you end up re-editing the clip you had finished.

Third, every language costs a render and an upload. Captions, by contrast, ride on top of the same cut: adding captions to a finished clip is a text job, not a re-edit — one timeline, many languages.

Auto-translate also has a captive audience, and it is not on your side. The viewer-side 'auto-translated captions' a platform offers switch on for the viewer, not for you, and they are the least reliable layer in the chain. The W3C's accessibility guidance on captions and subtitles is blunt about it: automatic captions are not sufficient, because a single missed word — 'not' — flips the meaning of the line.

The method: transcribe once, translate the captions

The trick is to separate the two things a caption carries: the text and the timing. Your cut already fixed the timing, so you translate the text and leave every timestamp where it is.

  1. Generate word-by-word captions on the finished clip. ClipFinish does this in the same pass that cuts the long video into vertical clips.
  2. Export the caption file (.srt or .vtt) and keep the original timings — do not re-time.
  3. Translate the lines of text, not the audio. Idioms first, then the literal pass.
  4. Re-wrap the translated lines, then re-burn: a translation is longer or shorter than the original, so the lines break in different places.

Step four holds most of the polish. CapCut's French guide for translated subtitles gives the numbers that keep a caption readable at speed: cut a line at 42 characters, use two lines maximum, and hold the reading speed at about 21 characters per second. A French line is routinely 15 to 20% longer than the English it came from, so a sentence that fit on one line now needs two — and if you do not re-wrap, the caption runs past the safe zone and off the foot of the screen. The same logic applies whatever the source: word-by-word captions are the raw material, and turning a long recording into shorts is where they start.

Bilingual or translated: pick the format

There are two ways to show a second language, and they are not interchangeable.

Translated captions replace the line: the viewer sees only the target language. This is the format for a real second market — a French audience watching an English-language clip gets French text and nothing else. It is the cleanest on screen and the least crowded, because it uses the space of a normal caption.

Bilingual captions stack both: the original and the translation, one above the other. UC Berkeley's guidance on multilingual video is explicit that for a bilingual audience you can put both language captions in the same file — and that each stays in the language actually spoken, so the two tracks form a translation pair, not a re-labelling. Bilingual captions are the format for language learners, for bilingual communities, and for clips where the original wording is itself the joke.

The trade-off is vertical space. Two stacked lines eat twice the height of one, and on a 9:16 clip the bottom of the frame is exactly where the platform's own buttons and captions sit. If you go bilingual, budget that space before you burn, or you will cover the face you cut the clip for.

What machine translation gets wrong on shorts

Machine translation is good enough to draft and not good enough to ship unread. On a clip it fails in three predictable places.

The three failures: idioms and humour, where a literal translation is correct and dead; community and game slang, whose vocabulary is in no dictionary the tool trusts; and names. A streamer's handle, a game, a sponsor — all get 'corrected' into something that does not exist. None of these is caught by a spell-check, because the output is spelled perfectly.

The honest limit of the tool matters here too. The ClipFinish clip pipeline turns a long recording into captioned, reframed vertical clips in a single pass, and its captions are word-by-word and already timed to the cut. What it does not do is translate: it writes the captions in the language actually spoken. The translation is the one step you take on the exported text, which is exactly why it is cheap — you are editing a subtitle file, not a timeline.

Questions clippers ask about translated captions

Do bilingual and translated captions need separate files?
Yes. A translated file holds one language; a bilingual file puts both lines in the same cue. Keep the timings identical between them so you can swap files without re-cutting.
Which format grows an audience faster?
Translated captions, for a market that speaks the target language — one line, full screen, nothing crowded. Bilingual captions are for a mixed or bilingual audience and for teaching; they double the vertical space for a smaller gain.
Can I just use the platform's auto-translate?
You can, but it runs viewer-side and stays unreliable, and the W3C's position is that automatic captions are not sufficient without a read-through. Your own translated file is the version you control.
How long does translating a clip take?
The transcription is free because you already have it; translating a thirty-second clip is a handful of lines. The work is the read-through and the re-wrap, not the machine pass.

One clip, two audiences, one cut. Put your clips in a second language by translating the text you already have: keep the timings, re-wrap the lines to the 42-character rule, and read every line before it ships. Start from a finished, captioned clip — the ClipFinish clip pipeline produces the word-by-word captions and the vertical cut in a single pass, and the caption file it exports is exactly what you translate.

You pick the moments. ClipFinish does the rest.

Drop your long video, tick the moments in the transcript, and get the whole batch back: framed, captioned, ready to post.

Try for free

10 free minutes of finished clips every month. No card. No watermark.