How many words fit in a 15-second video ad?

We noticed it in our own output before anyone told us: some UGC clips had the creator talking noticeably fast. Not gibberish, not wrong — just rushed, like someone reading a disclaimer. The video looked fine on paper and felt wrong in the ear.

Word count was one factor we could control before generation.

The equation nobody writes down

A short video ad has a fixed length. A script has a fixed number of words. To deliver all of those words within that length, the average speaking rate would need to be:

speaking rate = words ÷ seconds

A 60-word script in a 15-second clip requires four words per second if every word is spoken. That leaves little room for pauses. A generated clip may rush, omit words or change the wording instead of following that calculation exactly.

Speech-rate controls depend on the model. A shorter script gives the model more room for pauses; changing audio speed alone after generation can affect its alignment with the video. Review the finished speech and lip sync together.

Word count is a useful starting point.

What "natural" actually is

Conversational English runs at about 150 words per minute, or 2.5 words per second. An energetic creator might push to 3; a calm one sits nearer 2. For an ad — where every word has to land — 2.5 is the honest ceiling.

Other languages differ, and not because people talk faster or slower. They differ because words carry different amounts. Hindi words are longer than English ones on average, so a Hindi speaker at the same natural tempo produces fewer words per second. Tamil, Telugu and Kannada — agglutinative languages that pack whole phrases into single words — come out lower still.

Here are the rates Advinky uses, per language, and what they mean for common clip lengths:

These are deliberately conservative — they already leave room for the breath at the start and the pause at the end of a clip — and they are estimates we are refining against real transcripts of finished videos. A language not on the list gets the slow rate, never the English one.

What this looks like in a real script

Take a 10-second Hindi ad: budget 20 words. That is a hook, one reason, and a call to action that says the brand name. It is not a hook, three benefits, an objection, a testimonial and a CTA — that is a 30-second script wearing a 10-second label.

The instinct when a script feels thin is to add words. The better move is to add seconds — or to accept that one strong reason beats three rushed ones. Almost every ad that "needed" 60 words was really two ads.

How Advinky enforces it

Three things happen between your direction and the script you approve.

1. The scriptwriter is given the budget, not just the length. It is told the language's rate and the total word cap, and asked to decide the beats itself — how many, what each one does, how long each runs — with the same rule per beat: a two-second hook holds about five words in English.
2. The result is measured. Language models are bad at counting their own words, so we do not trust the count; we compute it. Only spoken lines count — visual directions do not.
3. If it is over, it is fitted. A rewrite pass shortens the wording — filler, hedging, repeated ideas — while keeping the hook, the claim, the brand name and the call to action. If it cannot fit cleanly, you see the original with a note, never a silent failure.

You approve the fitted script, then review the generated audio. A word budget helps with pacing; it cannot guarantee that every word is spoken accurately.

The one-line version

If you write video scripts anywhere, for anyone: seconds × 2.5 = your English word budget. Halve your ambitions and double your seconds before you add a single word.

Shape your next script