Words per second, beat length, where to cut a breath and the three-second idea rule — 15 pacing tricks that keep a UGC ad moving without sounding rushed.
Most UGC ads do not lose people because the message is wrong. They lose people because the message arrives too slowly, or arrives at a constant speed so even that the ear stops tracking it. Pacing is a craft with numbers attached, and these fifteen are the ones worth memorising.
A 30-second UGC ad wants roughly 75–90 words. Under that, the delivery drags and you get dead air the editor has to fill with B-roll. Over about 3.2 words per second the voice starts sounding like a disclaimer read. Before you generate anything, paste the script into a word counter and divide by your target duration. This one check catches more pacing problems than any amount of editing afterwards.
A short-form viewer can hold roughly one new idea per three seconds. That means a 30-second ad carries about ten ideas — and most drafts try to carry twenty. Number the distinct ideas in your script. If you have more than ten, you are not writing a tighter ad, you are writing two ads. Split it: one about the problem, one about the mechanism. Each will outperform the crammed original.
Spoken sentences longer than about twelve words force the listener to hold a clause open while waiting for the verb, and in a noisy feed they simply drop it. Break long sentences at the comma. 'I've had this jacket for two winters and it still looks new even though I wear it almost daily' becomes 'Two winters in this jacket. I wear it almost daily. Still looks new.' Same information, three landing points instead of one.
Uniform sentence length is what makes a script sound written rather than spoken. Real speech swings: a nine-word sentence, then a three-word one, then a fragment. Go through your draft and mark the word count of each sentence. If four in a row are within two words of each other, break one and merge another. The variance is what the ear registers as a person talking instead of copy being read.
Reading silently hides every pacing problem, because your inner voice skips the hard parts. Read the script aloud at normal conversational speed and time it. If you run 20% over your target duration, the AI voice will too and something will get compressed badly. Sixty seconds of reading saves a render you would otherwise throw away, and you will hear the tongue-twister lines that need rewriting before they reach the voice model. Do this on the draft Klip Kanvas generates too — the script is editable before you spend a render on it.
When a video is 4 seconds too long, most people delete a sentence — losing information. Instead, tighten the silences. There are usually 15–25 inter-sentence gaps in a 30-second read, and trimming each by 150–200ms recovers three to five seconds without removing a single idea. Leave the pauses that carry meaning and take the ones that are just the speaker resetting. The ad ends up denser and, oddly, feels less rushed than one where you cut content.
A 0.4-second beat in front of a key line makes the ear lean in; the same beat after it lets attention leak out. 'And the price...' [pause] '...nineteen dollars' lands. 'Nineteen dollars' [pause] does not. Mark the three most important lines in your script and insert a comma, an ellipsis or an explicit pause immediately before each one. This is the single cheapest way to make an AI voice sound like it means what it is saying.
'So', 'and then', 'basically', 'I mean', 'honestly' and 'you know' feel natural when you write them and cost you two to four seconds of runtime each across a script. A handful of them do add authenticity — keep one or two, ideally in the hook where informality helps. Delete the rest. A 90-word script usually contains 10–12 words of pure connective tissue, which is a full three seconds you can spend on the demo instead.
Compare 'What I really like about it is that it dries in ten minutes' with 'Ten minutes to dry. That's what got me.' The operative information is the same but the second version delivers it in the stressed position where the ear is listening hardest. Go line by line and ask: what is the one word this sentence exists to deliver, and how far into the sentence does it appear? If the answer is past word four, rewrite it.
Visual pacing and verbal pacing are separate levers. In the first 10 seconds, change what is on screen every 2–3 seconds — angle, scene, insert, zoom — because early attention is fragile and motion holds it. After 10 seconds, stretch cuts to 4–5 seconds. Viewers who stayed that long are evaluating, not scanning, and rapid cutting during the demo makes it feel like you are hiding something the product cannot do.
A reliable 30-second structure: 0–3s hook, 3–8s the problem stated concretely, 8–18s the demo or mechanism, 18–25s proof, 25–30s CTA. Write the beat boundaries first with a word budget for each — about 9 words for the hook, 15 for the problem, 30 for the demo, 20 for proof, 12 for the CTA. Writing to a budget is faster than writing then cutting, and the resulting script has an actual shape instead of a slow slide toward the CTA.
A global speed increase of 4–7% is essentially inaudible — pitch correction handles it — and it shaves a couple of seconds off a 30-second ad while preserving every idea and every pause ratio. Push past roughly 10% and the delivery starts to sound chipmunked and anxious. Use this as the last step when an ad is marginally over a placement limit, not as a substitute for writing tightly in the first place.
Prices, percentages and the final instruction are the two places where speed actively costs conversions. Drop to about 2 words per second for the price line and the CTA, and add a short beat on either side. Rushed numbers do not get remembered, and a rushed CTA reads as an apology for asking. Everything before those moments can run fast; those two moments should feel like the speaker deliberately slowed down to make sure you got it.
Fast, dense pacing sells impulse products; considered purchases need room to think. Below roughly $30 AOV, run tight — 3 words per second, cuts every 2 seconds, 20–25 second total. Above roughly $80, slow to 2.3–2.6 words per second, let scenes breathe for 4–6 seconds, and allow 40–45 seconds of runtime. The expensive product needs the viewer to feel considered, and pace is how that feeling is manufactured.
A retention graph with a smooth decline is a pacing problem, not a content problem — nothing in the ad ever changes gear. A sharp cliff at a specific second is a content problem at that beat. Find the timestamp of each cliff and look at what happens 1–2 seconds before it, because the drop registers slightly after the cause. Tighten or re-cut that beat and re-run. Doing this on your top three spenders is usually a better use of an afternoon than producing another new concept.
Pacing is the difference between a script people follow and one they merely hear. Count your words per second, cap ideas at one per three seconds, cut breaths rather than sentences, and slow down only for the number and the CTA. Then let the retention curve tell you which beat still drags.
Create your first AI UGC video ad in minutes — no filming, no actors, no editing.
Try Klip Kanvas freeThirteen tips for collapsing the gap between an idea and a live test — pre-built campaign shells, good-enough thresholds, parallel setup and killing perfectionism.
Fifteen tips on hook rate and hold rate: exact definitions, realistic benchmark ranges, the 2x2 diagnostic, retention-curve shapes and which number to fix first.
Paste a product link and Klip Kanvas writes the script, casts the creator and renders the ad — no filming, no actors, no editing.
