Why your dubbed audio does not line up with the video

Your dub probably isn't misaligned. It's a different length from your original, and that's a translation problem rather than a technical fault.

The same sentence takes a different amount of time to say in different languages. Translate a 12 second English line faithfully into German and you don't get a 12 second German line, you get something longer. Do that fifty times across a video and the gap accumulates, so the first minute sounds fine and the last minute is visibly behind.

That's why dragging the audio clip left on a timeline never fixes it. You can align the start or you can align the end. You can't align both, because the two files aren't the same length.

Why is the dubbed audio a different length?

Languages differ in how much time they need to carry the same meaning. Some pack information densely, some spread it out, and translation faithful to meaning is rarely faithful to duration.

German is the one English speakers notice first, because compound constructions and verb placement tend to produce longer spoken renditions of the same idea. Japanese behaves differently again, with a syllable structure and rate that has its own relationship to English timing. The direction and the size of the difference depend on the specific language pair, and on the register: technical narration behaves differently from conversational speech.

The effect compounds. A half-second of extra length on each of forty sentences is twenty seconds by the end of the video.

So a dubbed track arriving at a different length isn't a defect report. It's the expected output of translating speech, and any tool that returns a dub exactly matching your original duration has made a choice somewhere to get there, usually speeding up delivery, cutting content, or compressing pauses.

Is something in the pipeline re-timing my audio?

Not on our side, and this is worth ruling out because people assume it.

We submit your whole file and three things: the file, the target language, and one flag. That's the entire call. No start time, no end time, no speaker count, no segment splitting, no per-line trimming. Nothing in our code opens up your audio and moves pieces of it around.

Dubbing is also automatic all the way through, and nothing is edited by hand afterwards. Our support page states that in as many words, which means single lines can't be fixed on request. That's an unusual thing for a vendor to write down, and it's the honest framing for a page like this one: if the timing of one sentence is wrong, there's no human step where someone nudges it back.

How can I tell how much drift I actually have?

You can read it off your own invoice, which is a diagnostic most tools don't give you.

Credits are minutes, always rounded up, minimum one per file. The function is one line: max(1, math.ceil(seconds / 60)). A 61 second file costs 2.

The part that helps here is what happens after you confirm a job. We quote you a price from a duration we read off your file, then the dubbing service returns the duration it measured, and the two are reconciled. If theirs is longer you're charged the difference, if it's shorter the difference comes back. So the number you were finally charged reflects measured audio rather than our estimate.

Compare that against the length of your source file. A file you expected to bill as 5 minutes that settled at 6 is telling you something about the length of the output before you've even listened to it.

What actually helps?

Three things, and the first is the one people skip.

Give it clean source audio, with one speaker at a time. This is the advice on our support page and it's not a throwaway line. Overlapping speakers, heavy background music and cross-talk all degrade the transcription that everything downstream depends on, and a translation working from a muddled transcript produces timing that's worse than the language difference alone would explain.

Work in shorter segments if timing is critical. Drift accumulates over length, so a 90 second segment can't drift as far as a 40 minute one. For course modules or anything cut to visuals, dubbing section by section keeps each piece's drift small enough to absorb, at the cost of more jobs. Each job is one file and one target language, so this is a real tradeoff in credits and not just in effort.

Accept that the track won't be frame-identical. If your video has on-screen text appearing at 4:12 and the dubbed narration reaches that point at 4:19, no dubbing tool resolves that for you. Either the edit accommodates it or the content is authored with the dub in mind from the start. Content built for localization tends to leave breathing room between beats rather than cutting tight to speech.

Where this bites hardest

Two cases are worth reading up on if this is your situation. Course content is the obvious one, because e-learning modules are cut tightly to slides and screen recordings: our guide to dubbing e-learning from English to German covers a pair where length difference is pronounced. Internal communications are the other, and dubbing corporate video from English to Japanese covers a pair where the timing relationship is different again. The full set is on the use case index.

FAQ

Why is my dubbed audio longer than the original?

Because a faithful translation of a sentence rarely takes the same time to speak as the original did. The difference is per language pair and it accumulates across a video, so a track that starts aligned can finish noticeably behind.

Can you shorten the dub to match my video?

No. Dubbing runs automatically and nothing is edited by hand afterwards, so individual lines can't be trimmed or retimed on request. If your video needs an exact duration, the practical route is dubbing in shorter segments and accommodating the difference in your edit.

Does lip sync fix this?

Lip sync addresses whether mouth movements match the sounds, which is a different problem from whether the track is the right length. A lip-synced dub of a longer translation is still longer. Length is the constraint to solve first.

Should I split my video before dubbing it?

If timing against visuals matters, yes. Drift accumulates with length, so shorter pieces drift less and each one can be checked independently. Bear in mind each segment is a separate job with its own charge, and credits are rounded up per file, so ten short segments cost more than one long file of the same total length.