Reading speed on a six-second ad

Count characters against six seconds. Cut words. Do not shrink type to cram a paragraph onto a 6-second spot. Reading speed is how fast on-screen words arrive. Characters per second is the unit you can actually count: glyphs, spaces, punctuation, divided by how long the cue stays up. A 6-second bumper has a hard clock. The eye is slower than a fast VO. Matching captions 1:1 to the waveform is how the last word vanishes.

This is the 6-second math, as a procedure. It is not a general CPS style-guide reprint.

Count the characters you actually have

Six seconds is a real duration on this catalog: Veo 3.1 lists 4/6/8s. Bumper-style paid units are often six. The runtime is not a suggestion. Type that needs eight seconds of reading will not finish.

Formula, dull on purpose: characters in the cue (including spaces) ÷ seconds the cue is on screen = CPS. A 51-character line up for three seconds is 17 CPS. Stretch the same line to one second and it is 51, which the eye will not finish.

Broadcast and streaming houses have used adult bands around 160–180 words per minute, and Netflix-style timed-text guidance has long used a CPS cap in the high teens for adult programmes, with a higher ceiling as a hard max. Those figures are other people’s style guides. They are not a Versely switch. Do not invent a house CPS and print it as a product spec. What you ship is count versus hold on this file.

Procedure:

Lock the 6-second picture. If you are still extending, you do not have a clock.
Write every on-screen string that will appear in those six seconds: speech captions, overlay claims, legal, price, CTA. They all spend the eye’s budget. Captions are not the only type.
For each cue, count characters including spaces. All-caps and emoji spend the budget too.
For each cue, note in and out. If you have not captioned yet, time the VO with a stopwatch and guess holds — then you will still measure after the burn.
Compute CPS per cue. If a cue is up for 0.8s with 28 characters, that is 35 CPS. It will fail on a phone.
Sum the unique claims the mute viewer must finish. If that list is three sentences plus a URL plus a disclaimer, you do not have a 6-second ad. You have a 15-second script on a 6-second plate.

Mute autoplay is the test environment. Sound-on viewers get the VO for free. Sound-off viewers only get what you held long enough.

Cut the sentence. Do not shrink the type.

Small letters at 30 CPS are still unreadable. Shrinking is how teams “fit” a paragraph. The line is still too long. Now it is also too small.

Fix order, mandatory:

Delete words. The 6-second claim is one sentence. “Ships tomorrow” beats “Our logistics partner has confirmed that your order will typically ship on the next business day.”
Split only if you have time to show two cues. On a 6-second file, two cues means each needs a real hold, not 400ms each. Split at a clause, not in the middle of the SKU.
Hold the last line a beat after the mouth stops. Matching the out-point to the last waveform blip clips the last word. You may hold into a pack freeze. You may not hold past 6:00 — there is no 6:00. End the sentence earlier instead.
Drop a competing overlay. A legal line plus a karaoke hook plus a price starburst is three scripts. Legal still has to exist if you owe it; then the hook has to get shorter, not the disclaimer in 8pt.
Only then consider type size, and only to keep two lines readable, never to pack a third.

Forbidden fixes:

8pt burn so the paragraph fits in the safe area.
All-caps to “increase impact” (it increases character-equivalent load and reads slower).
DYNAMIC motion to “energise” a dense offer. Motion does not add seconds. It costs 2× and spends attention.
Speeding the VO to 1.4× so more words fit. The ear might follow. The eye will not. The mouth will look wrong if you lipsynced.

Latin-script captions commonly sit near 32–42 characters a line and at most two rows. A 90-character auto cue is what you get when the speaker did not pause. On a 6-second ad that cue is the whole spot. Split or cut before you burn.

Names, URLs, and product spellings blow the cap and must not hyphenate. If the SKU eats a line, the claim on the other line has to be tiny in word count, not in point size.

Burn, then measure on the 6-second file

ASR will follow the VO. Your job is to stop it following a VO that is already too fast.

Procedure:

Lay the VO you can stand at a human pace. If the read only fits at auctioneer speed, recut the copy before add captions to video.
Transcribe. Language code = spoken language. 165 codes. BASIC (21 looks, 1×) is the default on a 6-second paid unit. DYNAMIC (9 looks, 2×) is a test, not a way to fit more words.
Edit the cues: cut words, split clauses, extend out-points to the end of the file if needed so the last claim holds.
Place two lines max, off chrome, off the pack letters.
Preview the first seconds, then watch the whole six muted at phone size. You are not sampling a hook. The whole ad is the hook.
Recompute CPS on what you actually burned. If a cue still fails, cut again. Do not export.

Checkable mute test on a 6-second file:

Pause at 0:01. Is a claim already readable (overlay or first cue). Logo-only fails.
Play through once, muted, thumb ready. Can you finish every line before it leaves.
At 0:06, did the last word survive. If the file loops, the last cue must not collide with the first. A leftover line into the loop is a second script.

If you generated the 6-second hero on Veo 3.1 with native audio, the baked line may already be too long. Replacing with TTS (4 credits on Gemini 3.1 Flash TTS) plus a shorter sentence is allowed. Shrinking captions on a long bake is not.

Trim a video does not add reading time. Trimming a 8-second take to 6 throws words away. Do that on purpose in the script, not as a surprise in the burn.

Budget type against the rest of the 6s

Reading speed is not only captions. Everything on screen is a cue.

Six-second budget, as a working pattern — not a law:

0:00–0:01: one burned claim (overlay or first caption). Frame-one poster.
0:01–0:05: one demo, one spoken sentence that the burn can finish.
0:05–0:06: CTA or legal, not both stacked, unless legal is a persistent small line you already user-tested at phone size.

If legal must be on the ad, it is a persistent overlay with its own hold, and the speech caption cannot sit on it. That persistent line still has a character count. Tiny unreadable legal is not compliance.

Speaker IDs and [music] brackets almost never belong on a 6-second pack bumper. They spend CPS on cast.

Chapters do not belong. Alt text does not replace this burn. A transcript of a 15-second original pasted onto a 6-second cutdown is how you get 40 CPS.

Archive the character counts with the cut. Take two of the same bumper should not grow a clause because someone “had a better line.”

Cut the copy to the clock, then burn in Versely. Bring a 6-second keeper and a sentence you can finish on mute. Do not bring a paragraph and a smaller font.

FAQ

What CPS should I set in Versely?

There is no Versely CPS spec on this page. Count characters versus hold on the file. Other companies’ high-teens adult caps are their guides, not a switch in the product.

Can I keep a long VO if I make the captions one short overlay?

You can. Then the mute viewer gets the overlay and the unmute viewer gets a different script. On a 6-second ad, make them the same short sentence.

Will DYNAMIC karaoke help people read faster?

No. The accent moves. The line still has to stay up long enough to finish. Cut words.