Caption translator

Subtitles, not dubbing, and the difference is the point

There are two ways to make a video understandable in another language and they are not close substitutes. Dubbing replaces the voice, which changes the performance, costs a great deal more, and in the synthetic version usually sounds like it. Translated captions leave the original voice alone and put the meaning on screen.

For short video the captions route is almost always right, and not only because it is cheaper. A clip carries a lot of its weight in how something was said, and a synthetic voice reading a translation throws that away while adding an uncanny quality that reads as low effort. The original voice with translated words on screen keeps the thing that made the moment worth clipping.

What it does not do is make you understood by an audience that will not read. That is a real limitation and worth being clear about rather than discovering after you have translated forty clips.

Why a line at a time and not a word at a time

Captions here are normally timed per word, which is what makes them hold attention. Translation breaks that, and the honest thing is to explain why rather than pretend it does not happen.

Languages do not put the same idea in the same order. A word for word translation produces something that is either nonsense or has the words arriving in the wrong sequence, because the object of a sentence in one language may come before its verb in another. So a translated line is translated as a whole line and timed as a whole line, which means the highlight follows the sentence closely rather than exactly.

In practice this reads well and it is not identical to the original. The Studio says so at the point you choose a language rather than leaving you to notice. If word accurate highlighting matters more than the translation does, keep the original language captions.

The part most tools claiming a hundred languages have not done

A right to left script is not a font swap. Arabic letters change shape depending on what they sit next to and have to be joined; the line has to break in the correct direction; and the text has to be laid out right to left while any embedded numbers still run left to right. Get any of that wrong and the result is text that is reversed, or letters standing separately like a ransom note.

This is done properly here, with four Arabic faces built in, which is a real differentiator rather than a marketing line. Checking every tool people compare us against, several localise their own interface into Arabic and none of them document right to left caption rendering. Interface locale and caption rendering are different problems and only one of them shows up in your video.

If you publish in Arabic, the practical advice is to test any tool with one real sentence before committing, ours included. The failure is instantly visible once you know to look for it.

Translate the corrected transcript, not the raw one

Order matters more here than people expect. Automatic transcription mishears proper nouns, technical terms and brand names, and a translator handed a misheard word will translate the wrong word confidently. One error becomes two, and the second is much harder to spot because it is in a language you may not read.

So fix the transcript first, then translate. Retype any line that came back wrong, or paste the real words if you have a script, and only then choose the language. The original is kept, so you can always come back to it.

How to do it

Specifications

The numbers rather than the adjectives, including the limits. If one of these is a dealbreaker it is better learned here than after paying.

What people ask about this

Keep going