Localize video instead of translating it

Why dubbed video looks dubbed

There is a reason the phrase “it looks dubbed” exists, and it is not the quality of the voice recording. It is that the video was made for one language and then had another one put on top.

The mouth was shaped for different sounds. The sentence lengths were set by different grammar. The on-screen text was written for a different reader. A translation workflow can align all of that — that is what lip sync, timing shifts and subtitle timing are for — but alignment is a repair, and repairs show.

None of that is a failure of the translation. It is what happens when the language changes and the video does not. It is also why dubbing services and voice over translation put so much of their effort into alignment rather than into performance: the performance was already fixed by the original shoot, so the only lever left is how well the new words can be bent to fit it.

The difficulty there is inherited, not chosen. It exists because the video and the language were produced separately — so the only lever left is alignment. Produce them together and there is nothing left to align. That is the whole difference, and it is what makes a localized video look made for its market rather than adapted to it.

So the answer is not a better translation. It is a video that was built in the target language in the first place — same structure, same pacing logic, same argument, new language, and a picture that was made for those words.

What actually changes per market

Localize is often used as a synonym for translate, and the two are not the same thing. The difference shows up the moment you list what actually has to change.

Two rows in that list are the reason a translated video can be perfectly accurate and still land wrong: the language was localizable, the video was not.

This is what makes multilingual marketing harder than it sounds. You are not shipping one asset in five languages. You are shipping five versions of an idea, and the idea has to be re-expressed each time — which is a production question, not a translation question.

Ad localization is where this bites first, because ads carry more than language. An offer, a currency, a season and a claim all have to be right for the market, and all four of them can be wrong while the translation is perfect.

One master, every market

Video localization services usually split into two offers: a translation pipeline, or a reshoot in each market. The first is cheap and shows its seams. The second is clean and costs a production per language.

There is a third position, and it is the one this tool takes: the structure is the master, and the language is one of its variables.

One master. Every market is an instance of it. Which is what changes the economics: the second market is not a second production, and the fifth is not a fifth.

There is a second effect that shows up later and matters more. When a market is a production, teams stop adding markets — not because a new market would not pay off, but because each one adds a permanent production commitment. When a market is an instance, the calculation flips: the marginal market is cheap enough that the question becomes whether the market is worth having, which is the question you actually want to be answering.

Which markets first: the usual advice is to start with the largest market for your category. That is a reasonable default and a poor one to follow blindly, because the cost of a market is not the same for all of them. Two things make a market cheap to add — how much of the video survives, and how stable the market's commercial details are. A market whose offer changes every month is a market you will re-edit every month, which is fine when that is a segment change and expensive when it is a new localization project.

Per market, not per language

Two markets can share a language and still need different videos. The UK and the US both read English; the offer, the currency, the spelling and the reference points do not match. Treating localization as a language list misses that.

That matters most where the volume is: ecommerce localization is not five translations of one product video. It is the same catalogue, adapted to each market's norms, currency, season and shipping promise — and then kept current when any of those move.

The useful distinction is the last row of that comparison. A paragraph can be translated and dropped in. A person speaking on camera cannot — unless the video was built so that the speaking part is a variable.

That is worth saying plainly, because content localization has a reputation for being a spreadsheet exercise: a list of strings, a list of markets, a column you fill in. It is that, for text. Video has no column you fill in — the equivalent of that column is a new take, and whether you can afford one is the whole question.

What stays, what changes

This is what rebuilding means here, component by component.

The left column is the answer to what gets kept: structure, not pixels. The right column is the answer to what gets rebuilt: everything a viewer in that market would notice. Because the speaking parts are variables, the mouth is not a problem to solve — it is generated for the words it is saying.

There is a mechanical reason this works, and it is worth one sentence: because the video is built as independent segments rather than one flat file, changing the language re-times the rest. The timeline and the related elements follow on their own, so you never adjust a timeline by hand.

In practice that turns localization from a project into a setting. A project has a kickoff, a scope and a delivery date. A setting has a value you change when the market changes. The reason teams treat localization as a project is not that it should be one — it is that in a translation pipeline, the only way to change the language is to start another project.

What this looks like in practice

One master, five markets, no reshoot.

What BrandU does

Four things, in the order they matter when you are adding a market:

• It rebuilds the video in the target language rather than translating it. The script is written in the new language, so the voice-over is native rather than aligned, and the picture is generated for those words — which is the part an AI video translation workflow has to work around, because translation starts from a finished file.
• It keeps one structure as the master. Every market you add is an instance of it: the argument, the shot logic, the caption placement and the pacing carry over, and the language is the variable.
• It makes a new market an edit, not a production. Changing the language re-runs the affected segments rather than rebuilding the video, which is why the fifth market is not the fifth shoot.
• It keeps markets current. Prices, offers and seasons move; a change to one of them is a segment change, not a new localization project.
• It puts the effort into the message rather than the mechanics. Rewriting a line, changing an offer or switching a market is an edit in the script, so the hours go into what the video argues in that market — not into re-cutting, re-timing and re-exporting it.

How to run this

You do not have to start from a reference video. If you have one, that is the strongest path, because the structure is already proven. If you do not, the create screen also takes a one-sentence description or a template — three ways in, same editor afterwards.

• Find a video that is already working in the category. Paste its page link — no download, no direct file URL needed.
• Bring your brand and product information in once. It is reused across every market.
• Pick the markets. A market is not the same thing as a language — set the offer, currency and reference points for each.
• Generate each version. The structure stays; the language and the speaking parts are new.
• When a price or an offer moves, change that segment. Do not redo the localization.