Translate an English script into French, Spanish, German or Polish and it comes back longer. Usually 15 to 30% longer in words. That is normal, it is not the translator padding, and it is the most common reason a locked video ends up with audio that will not fit the picture.
The short answer
Budget for roughly 20% expansion when writing English that is going into Romance or Slavic languages, and write the English script short to absorb it. A 30 second slot needs about 75 English words. If the same script is going into Spanish, French and Polish, write it to 62 words instead. Fixing it at script stage costs nothing. Fixing it after the session costs a re-record.
The part most people get wrong is what happens next. A script that is 20% longer in words does not automatically produce audio that is 20% longer, because Spanish and French are spoken faster than English. But it does produce longer audio. If your English master runs 29 seconds in a 30 second slot, the Spanish version is going to be a problem, and you will find out about it in the booth, which is the worst possible place.
Why translation makes text longer
Three things drive it, and none of them are solved by hiring a better translator.
- Romance languages need more wordsEnglish will happily say “upload speed”. French needs “vitesse de téléchargement ascendant”. English compounds nouns by shoving them together and reaches for short Anglo-Saxon verbs where French and Spanish need a phrase.
- Grammar adds obligatory materialGendered articles, agreement, prepositions that English drops. “Car keys” becomes “les clés de la voiture”. Two words to five, and nothing extra has been said.
- Marketing and technical English is denseIt leans on jargon and abbreviation that other languages have to unpack. The more compressed your English, the worse the expansion.
German is the odd one out. It builds long compound nouns, so the word count can barely move while the character count runs away. That matters more for on-screen text and subtitles than it does for a German voice over.
Expansion and contraction by language
These are the working figures the localisation industry uses, and they line up with what we see across multi-language campaigns in practice.
| Language | Change from English | Effect on your timing |
|---|---|---|
| Spanish | +15% to +30% | Runs long |
| Portuguese | +15% to +30% | Runs long |
| Polish | +20% to +30% | Runs long |
| Arabic | +20% to +25% | Runs long |
| French | +15% to +20% | Runs long |
| Hindi | +15% to +35% | Runs long, wide range |
| German | +10% to +35% | Character count worse than word count |
| Italian | +10% to +25% | Runs long |
| Russian | +15% | Runs long |
| Czech | +10% | Slightly long |
| Greek | +10% | Slightly long |
| Norwegian | -5% to -10% | Runs short |
| Swedish | -10% | Runs short |
| Danish | -10% to -15% | Runs short |
| Korean | -10% to -15% | Runs short |
| Finnish | -25% to -30% | Runs well short |
| Japanese | -10% to -55% | Count characters, not words |
Two things to notice. They are ranges rather than fixed numbers, because subject matter changes everything. And the bottom of the table contracts rather than expands, which brings the opposite problem: a Finnish script that comes back 30% short leaves the voice artist stretching the read to fill picture, and that sounds every bit as wrong as rushing.
Japanese and Mandarin sit outside this logic entirely. Word count is close to meaningless in a language written without spaces, so those are counted in characters, and the count tells you very little about how long the audio will run.
A 20% longer script does not mean 20% longer audio. It means longer audio. Plan for it rather than hoping the faster language absorbs it.
Words are not seconds
This is the part that trips people up, and it is why a straight percentage applied to the word count will over-predict your runtime problem.
Languages are not spoken at the same speed. In 2019, Coupé, Oh, Dediu and Pellegrino published a study in Science Advances covering 17 languages. They found all of them transmitted information at roughly the same rate, about 39 bits per second, despite very different speech rates. Languages spoken quickly, like Spanish, carry less information per syllable. Languages spoken more slowly pack more into each one.
So a Spanish script that is 20% longer in words does not run 20% longer in the booth. Some of that expansion is absorbed by the faster delivery. Some of it, not all of it.
The dangerous version of this is the producer who reads “Spanish is spoken faster” and assumes the timing sorts itself out. It does not. The overrun is smaller than the word count suggests and it is still real.
Where the overrun actually hurts
It depends entirely on whether anything is locked.
- Broadcast and radio slots. A 30 second slot is 30 seconds. There is no negotiating with a traffic department. This is where expansion causes the most re-records, and why commercial work needs the timing solved at script stage.
- Video with locked picture. Animation hits, on-screen text, product reveals. If the English lands the product name at 00:18 and the German lands it at 00:22, the edit is wrong. Someone is re-cutting picture per language, and nobody budgeted for that.
- eLearning with slide timings. Modules built around fixed slide advances drift out across eight languages. Usually the most forgiving format, because slides can be re-timed, but only if someone owns that job.
- IVR and on-hold. Menu prompts answer to the caller’s patience rather than a timecode, but a 40% longer options menu genuinely costs you abandoned calls.
- Untimed narration. Corporate films, documentaries and audiobooks with no locked picture. Expansion here costs money rather than accuracy, because a longer script is more studio time.
Writing the English so the other languages fit
The English master is where this gets solved. A script destined for eight languages should be written differently from an English-only script, and it almost never is.
Work backwards from the target. Professional narration sits at roughly 150 words per minute, which is the figure we use for scheduling and runtime estimates. Take 20% off that for the English draft and the other languages have somewhere to go.
| Slot | English at 150 wpm | Write the English to | Headroom for +20% |
|---|---|---|---|
| 20 seconds | 50 words | 41 words | 9 words |
| 30 seconds | 75 words | 62 words | 13 words |
| 40 seconds | 100 words | 83 words | 17 words |
| 60 seconds | 150 words | 125 words | 25 words |
| 2 minutes | 300 words | 250 words | 50 words |
If German or Polish is in the set, take another 5% off. If the set is Nordic only, you have the opposite problem and can write closer to full length.
Everyone hates this at first
Writing to 62 words feels like writing a 25 second ad
It is. That is the point. The alternative is a Spanish version that runs 36 seconds and a conversation with your client about why the ad has to be re-approved.
What helps, and what makes it worse
Helps
- Brief for adaptation, not translation. A translator asked for accuracy gives you accuracy and a 36 second script. A translator told to hit 30 seconds makes choices. Say the target duration out loud.
- Send the video, not just the script. A translator working from a Word document cannot see that the product shot lands at 00:18. Give them the cut with burnt-in timecode.
- Leave air in the English edit. Two seconds of music-only between sections is two seconds of absorption in every other language.
- Give the word count target in the brief. Not the English word count. The maximum for the target language.
- Ask for a timed read before the session. Five minutes of checking beats a re-record.
Makes it worse
- Fixing it with a faster read. An artist can tighten a read by around 10%. Past that it stops sounding like a person and starts sounding like terms and conditions.
- Filling every gap in the English master. No air means no slack, and no slack means the problem lands in the booth.
- Locking English first. Once the English is signed off it becomes the constraint everything else bends around. Lock the timing brief first.
- Translating on word count alone. Word count is a proxy for duration, not a substitute for it.
- Assuming the faster language absorbs it. It absorbs some of it.
If you are putting a brief together, how to brief a voiceover agency covers what to include so the timing question does not come up late. One campaign, every language covers running the same message across markets without eight separate schedules.
Frequently asked questions
How much longer does English get when translated into French or Spanish?
Typically 15 to 30% longer in words. French sits at the lower end, around 15 to 20%. Spanish, Portuguese and Polish run higher, up to 30%. Ranges rather than fixed figures, because the subject matter changes the result considerably.
Does a 20% longer script mean 20% longer audio?
No. Spanish and French are spoken with more syllables per second than English, so some of the word expansion is absorbed by the faster delivery. The runtime overrun is real but smaller than the word count suggests. Plan for it rather than assuming it disappears.
Can you just speed up the foreign language read to make it fit?
Up to a point. A skilled artist can tighten a read by around 10% without it sounding rushed. Beyond that it audibly degrades, and on a 20 to 30% overrun you cannot get there with delivery alone. The script has to change.
Does text expansion affect subtitles the same way?
Same cause, different constraint. Subtitles are limited by reading speed and characters per line, not by how fast someone can talk. Languages that expand in characters, German especially, hit the subtitle limit long before they hit a voice over limit.
What about Japanese, Mandarin and Arabic?
Arabic expands like a Romance language, around 20 to 25%, and reads right to left, which matters for any on-screen text. Japanese and Mandarin have to be counted in characters, and the relationship between character count and runtime is loose enough that a timed read is the only reliable check.
Who pays when the translated script does not fit?
It depends who was asked to solve it. A translation commissioned with no duration brief was delivered as specified. Where we are casting and recording, we flag a script that will not fit before it reaches the booth, because a re-record helps nobody. Worth agreeing before you commission rather than after.
Does expansion change what the project costs?
For untimed work, yes. A 25% longer script is more studio time and more edit. For timed work the cost usually lands in the adaptation round rather than the recording, which is why solving it at script stage is cheaper as well as better.
Send us the script and the target languages
We will tell you whether it fits the slot before anything is booked, and come back with a shortlist of native voices. Usually within a few hours.
