New AI models in AI Studio

Try Now
PRIZMAD

Google Ads Now Dubs Video Ads Into 33 Languages

Google's Asset Studio dubs video ad creative into 33 languages, free for select advertisers. What a dub fixes, what it leaves to you, and the checklist.

Prizmad Team5 min read
Flat illustration of one large phone screen with a presenter, surrounded by three smaller phone screens and globe, speech-bubble and soundwave icons
On this page

Google Ads is rolling out Video Ads Dubbing in Asset Studio — an AI feature that translates and dubs existing video ad creative into 33 languages and locales. Search Engine Land reported it on September 18, 2026, after Google Ads consultant Jyll Saskin Gales spotted the option in her account and shared it on LinkedIn. The feature is free for the advertisers who have it, availability is limited, and Google has published no announcement and no help page — the Asset Studio documentation still describes adding a voice-over, not dubbing.

That is worth reading precisely. This is not a keynote promise or a "coming soon" slide: it is a shipping capability that arrived in accounts without a press release, eleven months after the same tool was first spotted as an internal, access-denied experiment in October 2025.

The real signal: dubbing has moved into the ad account

Look at the three platforms side by side and the direction is unambiguous.

PlatformDubbing capabilityBuilt for paid ads
TikTokSymphony Creative Studio — translate and dub, voice cloning, lip-sync toggle, automatic AI label on exportYes — the export goes back into Ads Manager
MetaReels AI translation and dubbing for creators (since August 2025); AI dubbing announced for Advantage+ in October 2025Partly — the creator feature is organic-side
GoogleVideo Ads Dubbing in Asset Studio — 33 languages and locales, voice track only, free for select advertisersYes — it lives inside the account that buys the media

For an ecommerce advertiser the practical meaning is simple: translation is becoming a default of the ad platform, not a line item in the production budget. The interesting question stopped being "can I afford a German version" and became "which versions should exist, and who checks them."

What a dub actually replaces — and what it does not

A dub replaces the audio track. Everything that is burned into the picture stays in the language of the original shoot: on-screen text, product packshots, subtitles, legal lines, disclaimers, price overlays. If your winning ad opens with a caption card that says "3 for 2 this week", the German viewer hears German and reads English.

Google has also not said what is happening under the hood. Search Engine Land's open questions are the right ones: whether the voice imitates the original speaker or is a generic synthetic voice, whether lips are synced, how much control the advertiser has over the translated script, and what gets an account into the select group. The performance figures in Google's own promotional material — an 86% increase in click-through rate and a 75% reduction in cost per click — are Google-provided examples, not an independent benchmark, and they should be treated as such.

So the honest summary of the news: the cheapest part of localization got cheaper, and the parts that decide whether a localized ad converts — the script, the claims, the on-screen copy, the landing page — are still yours. Our localization playbook covers that layer market by market, including the dub-versus-re-avatar-versus-rebuild decision, and the platform section there now includes this Google rollout.

Why this changes the variant math, not the strategy

The reason dubbing matters more than it looks is frequency. A paid creative that has been running at high frequency for weeks is tired; a localized version of the same structure reads as fresh inventory to both the viewer and the delivery algorithm, because each platform evaluates engagement on the localized asset. If variants are free to produce, the rational move is to localize the hooks that are already winning rather than to localize everything on spec.

That is also exactly how a generated pipeline behaves, which is where the economics get concrete. Prizmad renders a localized variant as a new render, not a re-dub of a finished file: the script, the voiceover and the captions are produced in the target language in the same pass as the visuals, so nothing is left burned into the picture in the wrong language. Token pricing is flat and published — ≈1,000 tokens per finished ad, on plans of $39/mo for 3,000 tokens (≈3 ads), $99/mo for 8,000 (≈8) and $249/mo for 35,000 (≈35), with full commercial rights on every plan. On Starter that is about $12 per finished ad, which is the number that decides how many markets you test this quarter. The wider cost picture across human, studio, agency and AI routes is in the AI video ads cost breakdown.

The five checks before a dubbed ad runs

  1. Can you actually serve the market? Currency, shipping terms, returns, support hours, and a landing page in the same language. A localized ad pointing at an English checkout converts worse than an English ad pointing at an English one.
  2. Transcreate the hook, don't translate the sentence. A hook that works on "throws away your money" logic often fails on "wastes your time" logic. Change the eight to twelve words the ad lives on, keep the structure.
  3. Re-cut the parts a dub can't touch. Captions, on-screen text, packshots with English labels, price overlays, legal lines. This is the work that stays, and it is why "free dubbing" is not "free localization."
  4. Check claims and disclosure per market. The EU AI Act's transparency obligation for deepfake-class content applies from August 2, 2026; New York's synthetic performer law (in force since June 9, 2026) requires conspicuous disclosure for AI-generated performances but carves out AI used solely to translate a human performer's own speech; platform labels still apply in every language. Our AI ad disclosure guide walks the platform-by-platform mechanics.
  5. Keep one naming convention across markets. If the German variant is not named like the English one, you cannot attribute the result to the market rather than to the asset. Localization only pays if the measurement survives it.

The honest caveats

  • It is limited and undocumented. Select advertisers only, no help page, no stated access criteria, no published behaviour on voice, lip-sync or translation control — and Google does not owe anyone continuity for an unannounced feature.
  • The headline numbers are vendor numbers. 86% and 75% come from Google's own promotional material, not from an independent test on a comparable account.
  • A dub is one of three routes, not the answer. Single-speaker, low-cultural-friction creative dubs well; when the presenter's credibility is market-specific, re-avatar; when the claims or packaging differ for compliance reasons, rebuild the affected beats.
  • Nothing here replaces a localized offer. The audio was never the bottleneck for a market you cannot ship to.

The pattern across six months of platform updates is consistent: the render layer keeps getting cheaper and more automatic, and the differentiation moves further into the script, the offer and the market-specific detail. That is good news for a team that can produce variants quickly, and neutral for everyone else.

Start with the market you can already serve: turn a product URL into your first ad, then localize the winner rather than the whole library. The full market-by-market workflow, disclosure rules and cost math are in the localization playbook.

Generate Your First Ad in 5 Minutes

Paste a product URL. Prizmad writes the script, picks the avatar, renders the voiceover with lip-sync, adds subtitles and music, and ships a TikTok / Meta / YouTube-ready mp4. No camera, no editor.