How to translate a video into another language
Upload the video to Vizard Agent and name the target language. Vizard Agent transcribes the original, translates it, generates a dub in a voice that matches the original speaker and adds subtitles. What it cannot change is arithmetic: a translated sentence is rarely the same length, so timing is where this needs your attention.
What is the short version?
Three moves, and the third one is not optional. Vizard Agent handles transcription, translation, voice matching and subtitles as a single request, which removes the four separate vendors this normally involves — but the fit between new audio and existing picture is something only watching it will tell you.
- Go to Vizard Agent and upload the video.
- Name the language, and say whether you want dub, subtitles or both.
- Watch it through once and fix anywhere the timing slips.
The rest of this page is what goes into each of those, and where the approach stops working.
What do you need before you start?
The video, and the target language. Because Vizard Agent works from what it hears, the audio decides the quality of everything downstream — a translation can only be as good as the transcript it came from, and a transcript is only as good as the recording.
- Clean speech. Music under dialogue, room echo and crosstalk all raise the transcription error rate, and every error propagates through the translation into the dub.
- One clear speaker at a time. Overlapping speech is hard to attribute, and a dub that puts the wrong words in the wrong mouth is worse than no dub.
- The language, named exactly. "Spanish" leaves a choice between Latin American and European; say which if it matters to your audience.
- A plan for on-screen text. Titles burned into the picture are pixels, not audio, and dubbing does not touch them. If your video has them, they stay in the original language unless you add a new graphics layer over the top as a separate job.
What do you type into Vizard Agent?
Vizard Agent ships this as a template, and its prompt covers the common case in one line: the language, and what you want produced. The language field takes plain words, so "Latin American Spanish" is as valid an answer as "Spanish" — worth spelling out when the variant matters to your audience. Fill in the brackets and send it with the video attached:
Prompt
Three variants worth keeping:
When you need several languages
When subtitles are enough
When you want to check the wording first
What are the steps inside Vizard Agent?
Three, and there is no hand-off between them. Transcription, translation, voice generation and subtitle timing normally involve three or four separate tools with a file passed between each; in Vizard Agent they are one request, which is why the timing check at the end matters more than any setting at the start.
- Upload the video to the Vizard Agent chat with the prompt above.
- Pick the tier that matches the job. For anything customer-facing, the product recommends Max — the free Flash tier is built for quick, simple edits.
- Watch the whole thing once. Not scrub it. Timing drift shows up in the middle of a sentence, and skipping through is exactly how it gets missed.
What does the result look like?
A video in the target language, dubbed in a voice close to the original speaker's, with subtitles. Vizard Agent keeps the transcript and the translation in the project, so correcting one wrong term and re-generating does not mean starting from the video again.
The picture comes back in the format the destination needs rather than whatever the source happened to be — a 4K master is delivered at 1080, 60fps comes back at 30, and footage bound for a vertical platform is reframed. If you need the original format preserved exactly, say so.
Cost is the thing to plan for here. Dubbing is billed by the minute of finished video — on the order of 200 credits per minute — which makes it the most expensive job on this list. Short clips stay cheap (the median translate-and-dub project runs about 33 credits, with the middle half between 12 and 197), but a ten-minute video is a different order of spend. The median run takes about 31 minutes, with the middle half between 14 and 46.
When does this not work well?
Dubbing runs into one structural problem that no tool escapes: languages take different amounts of time to say the same thing, so new audio never lands exactly where the old audio did. Vizard Agent absorbs most of that by adjusting the wording, but there is a point past which something has to give. Seven cases are worth knowing before you commit:
- Dense, fast speech. A rapid English sentence can become a German one half again as long. Something has to give: the pacing, the wording, or lip sync.
- Tight lip sync on close-ups. A wide shot forgives a lot; a close-up of a mouth does not. If the speaker fills the frame, expect visible mismatch.
- Music and speech on one track. Where the music sits on its own stem it comes through untouched. Where it is mixed into the same track as the speech, replacing the voice can affect it — worth checking on a heavily scored piece.
- Idioms and humour. A joke that works on a pun does not survive translation. Vizard Agent translates the meaning; if the meaning was the wordplay, it is gone.
- Names, brands and technical terms. These are what translation gets wrong most visibly. You can ask for specific terms to be left in the original language, but treat it as a request rather than a setting — check them in the result.
- On-screen text is not translated. This is a hard limit, not a degradation: dubbing replaces the audio and leaves the picture alone, so burned-in titles and lower thirds stay in the original language. Covering them means laying new motion graphics over the top, which is a separate job.
- Regulated or legal content. A translated compliance line is a new compliance question in that market. Treat the output as a draft for a native reviewer.
For anything where a mistranslation carries real cost — legal, medical, a market launch — have a native speaker read the transcript before you publish. That is not a limitation of this approach so much as the honest bar for translation generally.
How do you fix a result that came back wrong?
Reply in the same Vizard Agent conversation instead of starting over. The transcript and the translation are still there, so you are not charged again for work already done — only for what the correction itself costs. How much re-renders depends on what you change: a reworded line is cheap, a different voice is a fresh dub. Point at the moment and say what is wrong.
| What you see | What to say next |
|---|---|
| The dub runs past the shot | "Tighten the wording from 00:40 to 00:55 so it fits the picture" |
| A term is translated that should not be | "Keep the product name in English everywhere" |
| The voice does not sound like the speaker | "Use a lower, slower voice — closer to the original" |
| Subtitles are too long to read | "Two lines maximum, and hold each one at least a second" |
| The register is wrong | "Make it formal — this is for a corporate audience, not social" |
How does Vizard Agent compare to the usual localisation route?
The usual route is a chain: transcription vendor, translator, voice talent or a text-to-speech tool, then an editor to fit the new audio to the picture. Each hand-off costs a day and a file, and a single corrected term means walking the whole chain again.
| Vizard Agent | Transcribe → translate → record → edit | |
|---|---|---|
| Vendors involved | One request | Three or four |
| Time to a first version | About 31 min (half fall in 14–46) | Days to weeks |
| Cost of a term correction | One sentence | Re-record and re-edit |
| Several languages | One run each, no new upload | Multiply everything |
| Native-speaker review | Still yours to do | Still yours to do |
Neither replaces the other. Vizard Agent wins on speed, on cost and on doing ten languages instead of one; a human translator and a real voice actor still win when the writing itself is the product.
Common questions
Will the dubbed voice actually sound like the original speaker?
Vizard Agent matches the voice to the original — pitch, pace and general character — rather than cloning it exactly. It reads as the same kind of person speaking that language, which is what most viewers register.
Can I get subtitles without replacing the audio?
Yes — tell Vizard Agent explicitly: "subtitles only, leave the original audio". It is the safer choice for anything where a mistranslation matters, because your viewers can still hear the original.
How many languages can I do?
As many as you need, but they run one language at a time rather than all in a single pass — ask for several and Vizard Agent works through them, returning a file per language. The real limit is practical: every extra language is another one somebody should be reviewing before it goes out.
Does it translate text on screen as well?
No. Dubbing replaces the audio and leaves the picture as it is, so anything burned into the frame stays in the original language. Getting those translated means adding a graphics layer over the top — a separate job, and one where a longer translated title may not fit the space the original filled.
What happens to background music?
Where the music sits on its own stem it is preserved. Where it is mixed into the same track as the speech, replacing the voice can affect it — a heavily scored piece is the case most worth checking before you publish.