Write a multilingual ad script by holding the offer, the structure and the proof constant across markets, then writing the hook and the voiceover natively in each language rather than translating them. Translation reliably breaks two things: the idiom in the first three seconds, which is where the money is, and the length, because German needs 15 to 30 percent more seconds than English to say the same thing. Everything else can and should be shared, which is what makes six languages a batch rather than six campaigns.
What changes per market and what stays
This is the list we work from. The right column is the reason, because the reasons are what tell you how to handle a market not on the list.
| Element | Changes per market | Stays constant | Why |
|---|---|---|---|
| Offer and price terms | Currency, price point, delivery promise | The structure of the offer | A different offer makes the markets incomparable in reporting |
| Hook | Almost always | Only the hook's job | Idiom, humour and directness norms do not survive translation |
| Script structure and timings | No | Hook, stakes, offer by halfway, proof, ask | Slot structure is what makes results comparable across markets |
| Word count | Yes, considerably | Duration of the ad | German needs a fifth fewer words to fill the same seconds |
| Proof | Sometimes | The type of proof | A UK price comparison may need a local equivalent to land |
| Footage and shot selection | Rarely | Usually the same library | Faces and settings travel better than language does |
| Captions | Yes, rewritten with the script | Placement, chunk size, safe areas | Chunking rules are physical, not linguistic |
| Voice | Yes, native speaker or native voice model | Pace and tone brief | An accented read costs credibility in the first second |
| CTA verb | Yes | The action being asked for | "Tap through" has different natural equivalents per language |
| Compliance and claims | Yes | The claim you are making | Price and health claim rules differ by market |
The pattern: structure and intent are shared, surface and length are local. Teams that get this wrong usually get it wrong in one of two directions, either translating everything (which kills the hooks) or rebuilding everything (which produces six campaigns nobody can compare).
Hook idiom is what translation kills
The first three seconds are the part most dependent on shared reference, which makes them the part least able to survive a literal crossing. Some examples of what happens.
"A daily flat white is twelve pounds a week" is a strong English hook because the flat white is a recognised unit of discretionary spending. In German, translated faithfully, it lands as an odd specific: the drink is less standard and the sentence reads like arithmetic. The native version compares against something recognised locally, keeps the number, and drops the drink.
"That is the whole pitch" is a piece of English understatement doing a lot of work: it signals confidence by refusing to sell. French copy that says the same words reads as abrupt rather than dry. Italian tends to want more warmth in the same slot, and German copy is more comfortable simply stating the specification, which is often the better hook there anyway.
Rhetorical questions are the clearest case of a norm difference. They are weak openers in English and weaker still in German, where a direct statement of the specification outperforms almost anything softer. Hindi ads carry a mixed register, with English product vocabulary inside Hindi sentences, and translating that into pure Hindi makes copy that no ad in the market sounds like.
The rule that follows: brief the hook's job, not the hook's words. "Name the weekly cost of the habit we replace, with a number, in under three seconds" is a brief that produces six good hooks. The English sentence itself produces one good hook and five translations of it.
Length drift, by language
The mechanical constraint. Roughly what we measure at advertising pace, as typical figures rather than published ones.
| Language | Words per second | Words in a 15s read | Same message versus English |
|---|---|---|---|
| English | 2.5 to 2.8 | 35 to 40 | Baseline |
| German | 2.0 to 2.3 | 28 to 32 | 15 to 30 percent longer in seconds |
| French | 2.6 to 2.9 | 36 to 42 | 10 to 20 percent longer |
| Spanish | 2.7 to 3.0 | 38 to 44 | 5 to 15 percent longer |
| Italian | 2.6 to 2.9 | 36 to 42 | 10 to 20 percent longer |
| Hindi | 2.3 to 2.6 | 32 to 38 | Level to 10 percent longer |
The failure mode is specific and quiet. A 38 word English script reads in about 14.1 seconds and fits. The German translation of it, often fewer words, reads in about 17 seconds, and the two seconds that get clipped are the last two, which contain the call to action. The ad runs, looks fine in the library, and asks for nothing.
The answer is to write to the duration in the target language, not to translate and then compress. Speeding up the audio is the wrong fix: it costs comprehension in exactly the language that needs it most, and it flattens the pause before the price. There is more detail on this in our breakdown of ad script word count by duration, including the per-duration word budgets.
How the six languages get produced
Genyad, our product, supports English, German, French, Spanish, Italian and Hindi. Scripts are written natively in the target language rather than translated from an English master, the AI voiceover is generated per language, the duration of that read is measured against the slot, and a script that overruns is rewritten shorter for that language rather than played faster. Captions are regenerated with the local script rather than subtitled from the English one.
What we do not do, plainly: we are not an avatar or dubbing tool, so there is no synthetic presenter delivering your six scripts and no lip-synced version of one performance across languages. If your requirement is one on-camera person speaking six languages, a dedicated avatar or translation product fits that better than we do. HeyGen, for example, is built around avatar and translation video, at $29 per month as publicly listed in August 2026, and prices move. Our approach assumes you already own footage and want six native ads made from it, which is a different job. How multilingual video ads get built from one library walks through it, and what localisation covers beyond translation is the wider list of per-market decisions.
Native review on the first batch only
The step teams skip and then regret. Before you scale, have one native speaker per market review the first batch, and give them a narrow brief so the review takes twenty minutes rather than a week.
What to ask them:
- Does the first sentence sound like an ad from here, or like an ad from somewhere else?
- Is anything unintentionally formal or informal? The pronoun choice is the usual offender in German, French and Hindi.
- Does the price read naturally, including how the currency and decimals are spoken?
- Is the CTA verb the one people actually use for this action?
- Anything that is technically correct and nobody says.
Question five catches the most. Do this for the first batch in each language, fix the brief rather than the individual scripts, and later batches inherit the fixes. Reviewing every batch forever is not affordable at the volumes that matter: our 2026 fatigue benchmark reports most creative being effectively dead within three weeks and brands shipping 15 to 50 variants a month seeing 3 to 5 times longer campaign lifespan than quarterly refreshers, and six languages multiply that workload.
One practical warning about reporting. Keep the offer and the structure identical across markets so that when Spanish outperforms German you can attribute it to the market or the hook rather than to a different ad. Six languages with six different offers is not a test, it is six anecdotes.
Frequently asked questions
Should ad scripts be translated or written natively?
Written natively, working from a shared brief that fixes the offer, structure and proof. Translation preserves meaning and loses the idiom in the first three seconds, which is the part carrying the performance.
What should stay the same across languages?
The offer structure, the slot timings, the type of proof, the footage in most cases, and the caption placement rules. Keeping these constant is what lets you compare markets instead of comparing six unrelated ads.
Why does a translated script no longer fit the same duration?
Because languages differ in how many seconds they need for one idea. German runs 15 to 30 percent longer than English in seconds even with a lower word count, so a translated 15 second script overruns and the call to action is what gets clipped.
Do I need a native speaker to review the ads?
For the first batch in each language, yes, and it is worth twenty minutes of their time. Fix the brief from their notes so later batches inherit the corrections rather than repeating the review indefinitely.