Vizard Agent

How to say whether you want subtitles or a dub

Last updated 2026-09-17 · 8 min read

Tell Vizard Agent whether you want the spoken audio replaced or the original voice kept with text added. Those are two separate pieces of work with different costs and different results, and "translate this video" describes both — so the wrong one gets built, delivered and then thrown away.

What is the short version?

Subtitling keeps the original performance and adds a reading task on top of it. Dubbing removes that performance altogether and replaces it with somebody else's. The two solve the same problem for quite different audiences, and nothing in the phrase "translate it" tells Vizard Agent which of them you meant.

  1. Tell Vizard Agent whether the voice changes or the text appears.
  2. If it is a dub, say explicitly that you want no subtitles.
  3. Say whether the source has text burnt into the picture.

What do you need before you start?

The video and a view about your audience. Vizard Agent will build either, but the choice is about who is watching — a dub reaches people who will not read, a subtitle keeps the speaker's own voice and is far cheaper to revise.

What do you type into Vizard Agent?

Use the words "replace the speech" or "add subtitles" when you brief Vizard Agent. Either of those phrases is completely unambiguous, and "translate this video" on its own is the instruction that produces the wrong deliverable about half the time.

Prompt

Replace the spoken Russian in this video with Ukrainian speech — I want a dub, not subtitles. Remove the Russian subtitles that are burnt into the picture as well, since they will contradict the new audio, and do not add any new subtitles on top.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the sequence on a real video where the first pass was subtitles and the second was the dub that had actually been wanted. Everything before step seven is work that a single sentence at the start would have saved.

  1. Transcribes the speech with exact timings.
  2. Finds where the existing burnt-in subtitles sit in the frame.
  3. Builds a translated subtitle track and renders it over the old text.
  4. Fixes the line breaks and the overlap with the burnt-in captions.
  5. Covers the original burnt-in text so none of it shows through.
  6. Verifies the new subtitles are in sync with the speech.
  7. Checks how the spoken audio can be replaced instead.
  8. Generates the dub in the target language.
  9. Strips every text track and burnt-in caption from the dubbed version.
  10. Checks the speech itself in the final file, not just the picture.

Step nine is the part people do not anticipate. A dub replaces what you hear and changes nothing you see, so the old subtitles sit there in the original language saying something the audio no longer says.

Step seven is where the job actually started. Six steps of subtitle work were correct, careful and not what was wanted — and that is the ordinary cost of the ambiguity.

Step ten matters because the picture can pass while the audio has not changed. Checking the delivered file's speech is a different test from checking its frames.

What does the result look like?

Whichever one you asked for, complete. A dub with the old captions gone and no new ones added, or a subtitle track with the speaker's own voice intact — and Vizard Agent confirms which it built by checking the audio as well as the frames.

When does this not work well?

Each of the two choices carries its own cost. A dub loses the speaker's own voice and will drift from the lip movements; subtitles ask the viewer to read, and they fail completely on anybody scrolling with the sound off and their eyes somewhere else on the screen.

How do you fix a result that came back wrong?

Say which of the two you actually meant. It takes one sentence to tell Vizard Agent, and that sentence is the entire difference between the two deliverables — everything else about the job, including the transcript, stays the same.

How does Vizard Agent compare to doing it yourself?

By hand the ambiguity is the same and the cost is higher. You subtitle the whole video, show it to somebody, and discover they meant a dub — except now the subtitling was your afternoon rather than a few minutes of processing.

By hand Vizard Agent
The ambiguity Same, discovered late Resolved in the brief
Burnt-in text Left contradicting the dub Removed as part of the job
The wrong version Hours lost Minutes lost
Checking The frames look right The audio is checked too
Changing your mind Start again The transcript is reused

Common questions

Which should I choose? A dub for viewers who will not read; subtitles to keep the speaker's own voice.

Can I have both? Yes, but say so. Otherwise Vizard Agent gives you one or the other.

Does a dub remove the old subtitles? It should, and Vizard Agent treats that as part of the job.

Is a dub more expensive? Usually, and it is harder to revise afterwards.

Will it match the lips? Approximately. Ask Vizard Agent for lip sync as well if it matters to you.

Can it keep the original voice underneath? Yes. Ask Vizard Agent to hold it quietly if you want the performance audible.

What about several speakers? Each needs its own voice. Tell Vizard Agent who they are.

Can I get a subtitle file as well? Yes. Ask Vizard Agent for one; it is worth having even alongside a dub.

What if the burnt-in text is everywhere? Then removing it is most of the work. Vizard Agent will say how much.

Does the edit change? No. Vizard Agent leaves the cut exactly as it is in both jobs.

Can I change my mind afterwards? Yes, and the transcript is reused, so the second version is cheaper.

Which is faster? Subtitles, considerably, and Vizard Agent can deliver them the same session.

What if I only want part of it dubbed? Say which part, and Vizard Agent will dub that section and subtitle the rest of it.

How do I check it worked? Listen to the delivered file rather than watching it. That is where a dub lives.