A carousel rarely flops because of the design — the problem is usually the structure: it isn't clear why you'd swipe past the third slide. Below is a working prompt pipeline that breaks any topic into slides with roles, checks each slide for grip, and converts the result into simple JSON that's ready for layout.
The idea: roles, not a fixed 10-slide template
Most carousel guides hand you a fixed scheme — "10 slides, here's the structure of each". The problem is that not every topic can carry 10 slides without filler. The approach that works is to give the model slide roles (hook, content, proof, cta) and let it decide for itself how many slides a particular topic needs, so the story isn't artificially stretched or squeezed.
Step 1 — breaking the topic into slides by role
The prompt below doesn't ask for "a carousel" — it assigns an editor role and a strict role scheme inside which the model works on its own.
You are an editor of carousel content for social media.
Here's the topic: [write the topic in one phrase].
Break it into slides using this scheme:
1. hook — a gripping opener with no lies and no gratuitous clickbait
2. 2-4 content slides — value or step-by-step
3. proof — if there's a real example or number, otherwise skip it
4. cta — what to do next
For each slide give: the role, a headline (up to 8 words),
2 body bullets (up to 90 characters each).
Don't force a particular number of slides — use
as many as the topic needs, but no more than 7 in total.
What to put in yourself: the topic in one phrase and, if you have one, a real example or number for the proof slide — don't invent them, it's better to skip that block than to insert an unverified figure.
Step 2 — checking the hook on every slide
A quirk of social platforms: part of your audience lands in the middle of a carousel from another post or from the recommendations feed — meaning they never saw the cover. If a content slide doesn't grip on its own, it loses those readers.
Here are the carousel slides: [paste the draft from step 1].
Check each content slide separately: if you removed
every slide before it, would the first line grip on its
own? If not — rewrite the slide's opening phrase so that
it works as a standalone mini-hook,
without losing the meaning of the rest of the text.
The typical mistake is chasing clickbait in every line. A hook should grip honestly: with concrete value or an unexpected angle, not with a promise the text below doesn't back up.
Step 3 — converting to JSON
Once the copy is ready, you can have it expressed in a structured form that drops into any carousel render script without manual markup.
Turn the final carousel copy into JSON:
{
"slides": [
{"role": "hook", "title": "...", "sub": "..."},
{"role": "content", "title": "...", "body": ["...", "..."]},
{"role": "cta", "title": "...", "sub": "...", "url": "..."}
]
}
Reply with valid JSON only, no explanations before or after.
From there the JSON goes into a layout template — change one bullet and rebuild the whole carousel, with no manual editing of the design in a graphics editor.
Full pre-publish checklist
- The hook doesn't lie: what the first slide promises is actually delivered later.
- Every content slide grips on its own, even without the slide before it.
- No more than 7 slides — if the topic doesn't fit, that's two carousels, not one stretched-out one.
- The proof slide either contains a real example or isn't there at all — no invented numbers.
- The CTA says what specifically to do next, not an abstract "follow me".
How we do it
We mark up our own content for qvib.pro on the same principle — slide roles plus a hook check on each one: structure and copy first, then conversion to JSON and rendering from a template. It's not a one-off trick but a repeatable process you can describe once for your niche and reuse for any new topic.
Related links
- Prompts for generating content: qvib.pro/arsenal/prompts/
- The vibe coding section: qvib.pro/arsenal/vibe-coding/
- How the qvib engine works: qvib.pro/