How to dub a review video into Vietnamese with subtitles

Last updated 2026-08-21 · 6 min read

What is the short version?

A dub that ignores the original timing drifts, and by the third minute the voice is describing something that left the screen a while ago. Getting it right means measuring how long each line has and fitting the translated read into it.

1. Go to Vizard Agent and upload the video.
2. Name the target language and the voice you want.
3. Ask for subtitles in the same language.

What do you need before you start?

The video and the voice description. Vizard Agent transcribes and translates and times everything itself, and the detail worth supplying is what the voice should sound like, because a review, a tutorial and a commercial all want a different read in any language.

• The video. Any length; this run was two and a half minutes.
• The target language. Vietnamese, Spanish, whatever the audience speaks.
• The voice. Male, female, energetic, calm.
• Subtitles. In the same language, almost always worth it.
• Old subtitles. Say if the original has burnt-in text to work around.

What do you type into Vizard Agent?

State the voice in the same sentence as the language. Vizard Agent auditions candidates against that description before it records anything, and a phrase like "a male review voice" narrows the search far more usefully than naming the target language on its own ever does.

Prompt

Variants worth knowing:

• Keep the sound effects. Say so and Vizard Agent preserves them under the new voice.
• Several languages. Ask for each one; the timing analysis is shared.
• Subtitles only. If you want the original voice kept, say that instead.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real Vietnamese dub. The two halves of it are quite different jobs: first fitting the translated voice to the original timing, and then fighting the subtitle box until the Vietnamese diacritics stopped being clipped off.

1. Probes the video and extracts the speech and subtitles from the original.
2. Extracts frames and checks where the original subtitle bar sits at the bottom.
3. Searches for suitable Vietnamese male voices and records a test line to judge the tone.
4. Lists every segment of the original dialogue and calculates the reading speed and duration each one needs.
5. Generates the dub in three parts and checks each part's duration against the original timings.
6. Adjusts the speed and length of each segment so the dub lands where the original did.
7. Searches and downloads a music bed, then measures the loudness of the voice and the music separately and normalises the read.
8. Generates the Vietnamese subtitle file, renders a five-second test and looks at a frame.
9. Rebuilds the subtitle style repeatedly — position, background, size, line breaks — checking real frames each time until the full text displays, then verifies the middle and end of the video before delivering.

Step 9 is the part people underestimate in any language with diacritics. A subtitle style that works in English clips the marks above and below Vietnamese vowels, and the only way to find that is to look at a rendered frame.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1024x576, H.264, 30fps, 149.6 seconds, AAC audio. Two and a half minutes, the original length exactly, with a Vietnamese voice track timed to the original segments, a music bed, the game sound effects retained and Vietnamese subtitles.

The duration matching the source to the frame is the point of the timing work. A dub that runs long has to either speed up or cut, and Vizard Agent solved it at the segment level instead so nothing on screen falls out of sync.

When does this not work well?

Translation carries meaning across a language boundary and it does not carry everything with it. Vizard Agent produces a natural-sounding read in the target language, and some things simply do not survive the trip regardless of how well the dub itself is made.

• Jokes and wordplay do not translate. A pun in the original becomes a plain sentence.
• Lip movement will not match. This is a dub over the original picture, not a lip-sync.
• Burnt-in text stays in the source language. Vizard Agent can cover it; it cannot translate pixels in place.
• A translated voice is not your voice. If your identity matters, a cloned voice is a separate decision.
• Some languages run longer. Fitting a longer language into the same slot means a faster read.

How do you fix a result that came back wrong?

Say the segment and the problem. Vizard Agent keeps the original transcript, the per-segment timings, the voice test, the generated dub parts and the subtitle files, so a re-record or a restyle happens against measurements it already made.

• "That term is wrong in Vietnamese." Corrected and that segment re-recorded.
• "The voice is too fast at the start." Re-timed against the original segment length.
• "Subtitles cover the on-screen text." Repositioned and checked on a frame.

How does Vizard Agent compare to doing it yourself?

By hand this is commissioning a translation, booking a voice, discovering that the finished read is forty seconds longer than the picture it has to fit, and then rebuilding the subtitle style three times because the accents keep getting clipped.

Common questions

Will it match the original length? Yes. This run delivered at exactly the source duration by timing each segment.

Can I choose the voice? Yes. Describe it and Vizard Agent auditions candidates before recording.

Does it keep the sound effects? Yes. Vizard Agent mixes them back under the new voice track.

What about the original subtitles? Vizard Agent measures where they sit and works around or over them.

Can I keep the original music? Yes. Vizard Agent can preserve the existing bed and sit the new voice on top of it, or search for a replacement if you prefer.

Can it do several languages? Yes. Ask for each, and the timing analysis carries across.

Does it lip-sync? No. This is a dub laid over the original picture. Lip-sync is a separate job with a different tool.

How does it stop the dub drifting? Vizard Agent calculates the reading speed each segment needs, then adjusts the length of each generated part so it lands on the original timing.

Will the subtitles be readable? Vizard Agent renders test frames and rebuilds the style until the full text displays, which matters most in languages with diacritics.

Does it check the whole video? Yes. Vizard Agent verified the subtitle display in the middle and at the end of the video on this run, not only in the opening seconds where a test render usually stops.