How to dub a video into another language with AI, no reshoot

TLDR: Dubbing a video no longer requires reshooting or hiring a voice actor per language. HeyGen translates while keeping your cloned voice and matching lips to the new audio, ElevenLabs covers cases where you only need the voice, and Descript gives you the clean transcript to start from. The technology delivers. What decides the result is the script translation and a native review before publishing.

This guide is for anyone with content that works in one language and a potential market in another: corporate training, product video, creator channels, internal comms in distributed companies. The good material is already recorded, and re-recording it per language is what made going international expensive.

Important warning upfront: cloning a person’s voice or likeness requires their explicit permission. This guide assumes you are dubbing your own content or have written authorization from whoever appears.

1. Choose which videos deserve dubbing

Dubbing the whole catalog is the expensive mistake here. The translated version inherits the original’s performance: a video nobody watches in English will not be watched in German either. Sensible selection looks at two things.

First, permanence: prioritize evergreen material (explanations, training, product demos) over what expires in weeks. Dubbing cost amortizes over the time the piece stays alive. Second, proven performance: dub what already works, not what you hope will.

And the market question that avoids wasted work: will you serve that language? Publishing video in Italian and not answering Italian comments projects abandonment, not expansion.

2. Prepare and translate the script with judgment

Automatic dubbing starts from the transcript, so the transcript rules. Descript gives you that clean base: it transcribes the video and lets you fix proper nouns, spoken numbers and jargon on the text, which is exactly where automatic recognition fails. It has a free plan with 60 minutes a month and paid from $16.

With correct text, translation needs human judgment at three points: proper and brand names (which should not be translated), numbers and units (whose format changes by country) and local expressions or wordplay (which need an equivalent, not a literal translation).

This step determines more than any other whether the result reads as international or as a shortcut.

Dubbing technology is solved. Unreviewed translation is what exposes it.

3. Dub with lip sync in HeyGen

HeyGen does the visible work: it translates video into over a hundred languages, and its free plan gives three videos a month to evaluate it. If the video shows nobody speaking, lip sync is surplus and ElevenLabs does better and cheaper. The division rule is the shot itself:

Quality is not equal across languages: major ones perform better. Test yours on the free plan before committing budget.

4. Run quality control before publishing

Quality control on a dub has one valid standard: someone who speaks the language natively reviews the whole thing. Not a read of the script, the complete piece, because problems appear in the interaction between text and pacing, not on paper.

What that reviewer looks for fits in a short list, and anything they find gets fixed by regenerating only the affected fragment, not the whole piece.

Once the pipeline is built, the marginal cost of each new language is low, and that is the real change: deciding which markets to enter stops being a budget question and becomes a strategic one. To take dubbed content to social, the clips from long video guide closes the loop, and multilingual subtitles have their own guide. The whole sector lives in AI for content and media.

Common mistakes

Dubbing without native review. The technology sounds good and that is what deceives: the error sits in the text, and someone who does not speak the language will not catch it until a customer does.

Cloning other people’s voices without permission. Beyond the legal problem, this is the use destroying trust in these tools. Your voice only, or explicit written permission.

Dubbing the entire catalog at once. Start with three pieces, measure whether the market responds and scale on data. The opposite is prepaying for an audience that may never appear.

Forgetting on-screen text. The audio ends up in Italian while the captions stay in English. If the video carries burned-in text, it has to be redone too or the piece looks half finished.

Frequently asked questions

How much does dubbing a video cost?

HeyGen’s free plan lets you dub three videos a month to evaluate. Paying, from $24 a month, and if you only need voice without lip sync, ElevenLabs from $5 covers it.

Can people tell the dubbing is automatic?

The technical side, barely: cloned voice and lip sync have reached a level that holds up. What always shows is unreviewed translation, especially on proper nouns and local expressions.

Can I dub a video with several people speaking?

Yes, though the result improves the cleaner the voice separation is. If the original recording has separate tracks per participant, the dub comes out considerably better than from a single mix.

Is it legal to dub a video with someone else’s voice?

Only with their explicit permission. Cloning someone’s voice without authorization carries legal consequences in most jurisdictions and is also the practice costing these tools their reputation.