AI dubbing for ads: localize one winning video ad into 10 languages (cost and lip-sync compared)

AI dubbing for ads means replacing the voice in a finished video ad with a translated one, in the speaker's own voice, so a winning ad can run in markets you never shot for. The voice track is the cheap part: about 38 Orisu credits per language for a 30-second ad. Lip-sync, which you need only when a face talks on camera, is where the cost and the quality risk live.

This playbook covers when dubbing is enough and when you need lip-sync, what each language costs, how the main tools compare, and a review step that catches the problems AI still makes.

Why dub ads instead of reshooting or subtitling?

Because people buy in their own language. In CSA Research's "Can't Read, Won't Buy" study of 8,709 consumers across 29 countries, 76% of online shoppers said they prefer to buy products with information in their own language, and 40% said they will never buy from websites in other languages (CSA Research).

Subtitles help, but most social video is watched in a feed that moves fast, and a hook in a language the viewer doesn't speak loses them in the first two seconds. Reshooting with a local presenter works and costs a production per market. Dubbing sits in between: the creative that already won stays the same, and only the voice changes.

Do you need lip-sync, or just a dubbed voice?

This is the decision that sets the budget, so make it per ad, not per campaign.

The mixed row is the common one for performance ads. Lip-sync only the seconds where a face is talking and leave the product shots alone. On a 30-second UGC ad that is often a third of the runtime, which cuts the lip-sync bill by about two thirds.

Some video models now generate speech and picture together in the target language. If you are producing the ad from scratch with AI, generating each language version natively can beat dubbing an English master. Happy Horse 1.1, for example, generates multilingual lip-synced speech in the same pass. For an ad you already have, dubbing is the faster route.

What does AI dubbing cost per language?

Costs below come from Orisu's model catalog on October 6, 2026, with dollars at the Pro plan rate of $0.0098 per credit. Each row is one language for a 30-second ad.

So ten languages for a 30-second ad comes to about 380 credits (roughly $3.70) with voice only. With budget lip-sync on the full 30 seconds it is about 2,630 credits (roughly $25.80). Lip-sync only the 10-second talking hook and the lip-sync part drops to a third of that.

If you call the vendors directly, ElevenLabs lists Dubbing v1 at $0.50 per minute without a watermark ($0.33 with one) on its API pricing page, and Sync Labs lists lipsync-2 at $0.05 per second and sync-3 at $0.1334 per second on its pricing page (both checked October 6, 2026). Our full per-second table, including the video models, is in the AI video model price index.

How do the AI dubbing tools compare?

Every number here is from the vendor's own site, checked on October 6, 2026. Where a vendor doesn't state something, we left it blank rather than guess.

The pattern: all-in-one translators (HeyGen, Rask, Perso) bundle translation, voice and lip-sync, and bill by minutes or credits. Component tools (ElevenLabs for voice, Sync Labs for lips) are cheaper per second and let you skip lip-sync where you don't need it. For ads, where most of the runtime often isn't a talking face, the component route usually costs less.

Meta also lets you run multi-language ads, where one ad carries different text, images or video per language, and it can auto-translate the copy (labelled "Automatically Translated"). That handles delivery. It doesn't create the dubbed video for you.

The AI dubbing workflow for one winning ad

1. Lock the master

Pick the version that won and finish it before you translate anything. Every change after this point multiplies across languages. Export it with the voice on its own track if you can; clean separation between voice and music gives every dubbing tool an easier job.

2. Translate the script and have a native speaker read it

Translate the script first, then dub from the approved text. Machine translation handles grammar well and still misses idioms, product names and claims. A health or price claim that is fine in English can read as a promise you can't keep in another market. A native speaker should read every line, especially the hook. Keep your brand's tone in the brief for the translation step, so the German ad sounds like the same brand rather than a phrasebook.

3. Dub the voice

Run the dubbing step per language with the speaker's voice kept. Check that timing holds: German and French often run longer than English, and a line that overruns the cut needs a shorter translation, not a faster voice.

4. Lip-sync only the talking shots

Split the ad into talking and non-talking segments and send only the talking ones through lip-sync. Review them at full size. Lip-sync errors (smeared teeth, a jaw that lags) hide at thumbnail size and show up on a phone screen.

5. Swap on-screen text and captions

On-screen text has to change too. Re-render text layers in the target language rather than asking an image or video model to draw the letters; our notes on text in AI images explain why. Burn in captions if your placements autoplay muted.

6. Review, label where required, ship

Watch every language version end to end before it runs. If a platform or market requires disclosure of AI-altered content, apply it here. Our guide to AI ad disclosure rules covers Meta, TikTok, Google and the EU.

Common mistakes

• Dubbing before the master is final. Every later edit then has to be redone in every language.
• Lip-syncing the whole ad. You pay for seconds where nobody is talking.
• Skipping the native read. A mispronounced product name in market six undoes the polish of everything else.
• Leaving English text on screen. A dubbed voice over an English headline looks like a mistake, because it is one.

The runnable version

In Orisu the dubbing workflow is a graph: the master video, a translate step that carries your brand voice, the ElevenLabs dubbing node, a lip-sync node for the talking segments, and a review gate. Each market is one language input. Only the language-dependent steps run again, so adding market eleven doesn't re-bill markets one to ten, and the cost shows on the node before you run it.

Start from the ad localization workflow, or open a template and dub your best ad into one new market first. One language is enough to see whether the hook survives translation.