A 15 second video ad holds roughly 35 to 40 words of English voiceover once you leave room for pauses and a held end card, which is about 2.5 words per second of speaking time. The same message in German usually needs a shorter word count and still runs longer in seconds, which is why a translated script that fitted 15 seconds in English overruns. The fix is to rewrite the script shorter in the target language rather than speed up the audio.
Words per second, by language
These are the working figures we measure at advertising pace, which is faster than audiobook narration and slower than a disclaimer. Treat them as typical rather than published: measure your own voice and adjust.
| Language | Words per second | Words that fit in 15s | Same message versus English | Why |
|---|---|---|---|---|
| English | 2.5 to 2.8 | 35 to 40 | Baseline | Short words, few inflections |
| German | 2.0 to 2.3 | 28 to 32 | 15 to 30 percent longer in seconds | Compound nouns, longer words, more syllables per idea |
| French | 2.6 to 2.9 | 36 to 42 | 10 to 20 percent longer | More words for the same idea, but each is quick |
| Spanish | 2.7 to 3.0 | 38 to 44 | 5 to 15 percent longer | Fast syllable rate offsets extra words |
| Italian | 2.6 to 2.9 | 36 to 42 | 10 to 20 percent longer | Similar to French, more vowel length |
| Hindi | 2.3 to 2.6 | 32 to 38 | Roughly level to 10 percent longer | Depends heavily on how much English vocabulary stays |
Two columns to read together: words per second and the comparison to English. A language can fit more words and still take longer, because what matters is seconds of audio for one complete idea, not word count. Spanish is the clearest case: more words, faster delivery, small net overrun.
The German problem, with numbers
Here is the failure in the form it actually arrives in. An English 15 second script:
"Month-end close took nine days. Six of them were spreadsheet reconciliation. Ledgerly pulls both systems into one view, free for two ledgers. Start with your two busiest."
That is 38 words. At 2.7 words per second it reads in about 14.1 seconds, which fits with a beat of silence before the end card.
Translate it faithfully into German and you get around 36 words, fewer than the English. Read it at 2.1 words per second and it takes about 17.1 seconds. You are 2.1 seconds over a 15 second slot, and the part that gets clipped is the last sentence, which is the call to action. The ad still plays. It simply stops asking for anything.
Now rewrite it natively in German for a 15 second read, which lands around 28 to 30 words: state the nine days, drop the second sentence's detail, keep the offer, keep the ask. That reads in about 13.8 seconds and the CTA survives.
The pattern generalises. Expect the German version of a script that fits an English 15 seconds to need roughly a fifth to a quarter fewer words. On a 30 second ad you have slack and the drift is survivable. At 6 seconds it decides whether the offer is in the ad at all.
Why rewriting beats speeding up the audio
The tempting fix is to run the voice at 1.15x and move on. We do not do that, for four reasons.
- Comprehension goes first in the language that needs it most. German consonant clusters and compound nouns are the parts that blur when you compress, so you lose the specific nouns and keep the filler.
- It removes the pauses, and pauses are where the offer lands. A script with no beat before the price reads as a list.
- It sounds cheap. Accelerated speech has a recognisable texture, and audiences read it as a hard sell before they parse a word of it.
- It hides the real problem. The script is too long for the format, and the format is not negotiable.
Rewriting shorter forces the useful decision: which sentence was carrying the least. In our experience that is almost always the second half of the stakes line, not the proof and not the offer.
There is one legitimate use of small tempo changes: trimming 200 to 300 milliseconds off a read that is marginally over, where nobody can hear the difference. Anything past about 5 percent is a rewrite.
Measure the read before you assemble the video
The order that avoids the whole problem: write, synthesise, measure, rewrite if over, then cut picture to the measured audio. Cutting picture first and then discovering the voiceover is 2 seconds long is how a good script becomes a rushed ad.
This is how Genyad, our product, handles it. Scripts are written natively in the target language rather than translated, the AI voiceover is generated, and the duration of that read is measured against the slot. If the read overruns, the script is rewritten shorter for that language and re-synthesised. We do not speed up the audio to make a long script fit. The six supported languages are English, German, French, Spanish, Italian and Hindi, and each one gets its own script rather than a translation of the English, which is the same reason the German version comes out shorter in words.
If you are doing it by hand, the equivalent workflow is: read the script aloud on a timer three times and take the slowest, or synthesise it once in whatever voice tool you use and read the file duration. Both are more accurate than counting words, because pauses and numbers ("twenty-two pounds fifty") take longer than their word count suggests.
Word counts by duration
Rounded targets for English, with the German figure alongside since that is the constraint that binds in a six-language set. Subtract roughly 10 percent from the English number if you are holding an end card for four seconds inside the runtime.
| Duration | English words | German words | Structure it supports |
|---|---|---|---|
| 6s | 14 to 17 | 11 to 13 | One claim and an instruction, no proof |
| 10s | 24 to 27 | 19 to 22 | Hook, offer, CTA |
| 15s | 35 to 40 | 28 to 32 | Hook, stakes, offer, one proof, CTA |
| 20s | 48 to 54 | 38 to 43 | Adds a second proof beat |
| 30s | 72 to 80 | 57 to 64 | Adds objection handling |
| 60s | 145 to 160 | 115 to 130 | Full direct response, rarely worth it in feed |
Numbers spoken aloud are the most common reason a script that counted correctly still overruns. "€29 for 15 credits" is five words and about 2.2 seconds.
Why this matters at volume
If you run one campaign in one language, you can fix the timing by ear. Run six languages against a batch of 8 to 20 live variations and the German overrun happens on every variation at once, silently, until someone notices the ads with no call to action.
It also compounds with refresh rate. Our 2026 fatigue benchmark reports CTR declining 15 to 20 percent in a creative's first two weeks and most creative being effectively dead within three weeks, so those six-language batches get rebuilt often. A per-language word budget in the brief is cheaper than catching it in review each time. What localisation covers beyond translation and how we run multilingual video ads go through the rest of the per-market decisions, including which parts of a script should not survive the crossing at all.
Frequently asked questions
How many words are in a 15 second ad script?
About 35 to 40 words of English voiceover, which is roughly 2.5 words per second once you allow for pauses and a held end card. In German the same slot takes 28 to 32 words, because German runs longer in seconds for the same idea.
How many words fit in a 30 second ad?
Around 72 to 80 in English and 57 to 64 in German. Give the extra seconds to proof and objection handling rather than to a longer hook, which does not improve with length.
Why does my German ad script run over when the English fitted?
German needs more syllables and longer compound words to carry the same idea, so a faithful translation runs 15 to 30 percent longer in seconds even when the word count is lower. Rewrite it natively to the target duration instead of translating the English.
Should I speed up the voiceover to make a script fit?
Only for the last 200 to 300 milliseconds, where it is inaudible. Beyond about 5 percent, compression costs comprehension, flattens the pauses that sell the offer, and sounds like a hard sell. Cut a sentence instead.