Carousel modes, explained: per-slide vs single-repeated, slide caps, and per-slide overlays

If you opened the campaign builder a month ago, every carousel was the same shape: AI generates one image per slide, you got 5-10 slides, you posted it. Done.

That's not how carousels work in the tool anymore. Two modes, a slider for slide count, per-platform caps, and per-slide text overlays — all surfaced on the campaign builder + review step. Picking the right combination for a specific post is now a real decision. Here's how to think about it.

Two modes: per-slide and single-repeated

Every carousel post has a mode that determines how slides relate to each other.

Per-slide is the classic carousel: each slide has its own image and its own caption-overlay text. Five slides = five distinct images. Use this when each slide is a different beat — like a "5 tropes I used to hate" carousel where each slide is its own trope, or a launch teaser where slide 1 is the cover, slides 2-4 are atmosphere, slide 5 is the CTA.

Single-repeated uses one image across all slides — the same image is the canvas for different text overlays per slide. Five slides = same image, five different text layers. Use this when the visual is the unifying element and the text is doing the storytelling. Good for: book quotes (one cover image with five quotes overlaid), an "if you liked X, try Y" recommendation list laid over a single aesthetic, a step-by-step craft tip where each slide adds a step on top of the same illustration.

Pick the mode in the campaign builder before generation; the review step lets you override per-day if you change your mind.

Slide count: 3 to 25, on a slider

The slide-count slider on the campaign builder lets you pick anywhere from 3 to 25 slides up front. Defaults to 5 because that's what most authors land on, but you can drag it.

Practical guidance:

• 3-5 slides — for skim-friendly content. Hook, one or two body slides, CTA. Highest completion rate.
• 5-7 slides — the sweet spot for most narrative carousels. Hook, 3-5 body beats, payoff, CTA.
• 8-10 slides — instructional or list content where each item needs its own slide. Drop-off increases past slide 8 — make slides 1-3 strong enough to hook even if not everyone makes it to the end.
• 10+ — only for genuinely list-heavy content (every back-of-the-book trope explained, every chapter title with a one-line teaser, etc.). Most carousels don't need this.
• 25 — the cap. Useful for TikTok photo carousels where the algorithm rewards depth.

If AI generates fewer slides than you asked for, the tool pads up to your requested count. If it generates more, it trims. Either way you get exactly the slide count you picked.

Per-platform slide caps

Different platforms have different maximum carousel sizes. The tool enforces these at schedule time so you don't ship a 10-slide carousel to a platform that only allows 4.

If you're generating a 10-slide carousel and posting to all of the above, the composer warns you before scheduling — Bluesky and X will need a trimmed version. You can either drop those platforms from this carousel or split into a separate per-platform plan.

(Per-platform carousel splitting where the tool generates different slide counts per platform from one campaign concept is on the roadmap. For now it's pick a count and either trim manually or skip the platforms that can't handle it.)

Per-slide text overlays

Each slide in a carousel can have an optional text overlay — typewritten text composited onto the image at FFmpeg-render time. Authors use this for:

• Hook text on slide 1 (large type, short sentence, stops the scroll)
• Quote attribution on quote-card slides
• Step numbers + headings for instructional content
• Book title + tagline on a cover-reveal slide

The overlay editor on the campaign review step gives you per-slide control: text content, font size, position, color, and a backdrop for legibility. It's lightweight by design — five fonts, three positions, three colors — because we're optimizing for "post something good in five minutes" not "design pixel-perfect typography in an hour." For pixel-perfect work, design in Canva and use Upload-Your-Own (next section).

Upload-Your-Own per slide

Every slide tile on the review step has an Upload button. Drop in your own image; it replaces the AI-generated slide for just that slide. The caption stays.

Use cases:

• The first slide is your actual book cover — generate the rest, upload yours for slide 1
• One slide needs a real photo (you, your workspace, an in-store shot) the AI can't produce
• A specific aesthetic you've already designed in Canva — paste the export in
• Mid-carousel content moment that needs to be exactly right (an author photo, a fan's repost screenshot, a chart)

Mix-and-match works. A 7-slide carousel with slide 1 uploaded, slides 2-5 AI-generated per-slide, slide 6 uploaded, slide 7 single-repeated text-overlay is fine. The tool handles the coordination.

Cross-platform behavior in single-repeated mode

Single-repeated mode has a quirk worth knowing: when you post a single-repeated carousel to a platform with fewer allowed slides, the result still works because the image is the same — the platform just shows fewer text overlays.

Example: a 10-slide single-repeated carousel for Instagram + a parallel 4-slide auto-trim for X. The Instagram audience sees the full storytelling arc; X sees just the hook + payoff slides (slides 1, 4, 7, 10 by default — you can override which subset ships per platform).

This is not true for per-slide mode, where dropping slides means losing distinct content. Single-repeated degrades more gracefully across platforms.

What AI is good at vs needs review

Per-slide carousels lean on AI image generation, which is strong at:

• Aesthetic backgrounds (warm-light coffee shop, moody coastal cliff, abstract botanical pattern)
• Generic scene illustrations matching your imagery style guide
• Cover-adjacent art (atmosphere of your book without showing characters AI might get wrong)

AI image generation is weak at:

• Specific characters with consistent appearance across slides (use Upload-Your-Own with your own art for this)
• Text inside the image (use the per-slide overlay editor instead)
• Hands, eyes, fingers, complex architecture (regenerate, or swap to a different provider via Regenerate-with)

Single-repeated mode sidesteps most of these issues because the image is one canvas you control once.

Common mistakes

1. Picking 10 slides for a topic that needs 5. More slides isn't more value. Honest 5 beats padded 10 every time.

2. Generating per-slide art when single-repeated would be cleaner. A book quote carousel is almost always single-repeated. Five different background images for five quotes feels chaotic.

3. Forgetting per-platform caps. Your 10-slide Instagram carousel does not become a 10-slide X carousel. Check the warning, decide whether to skip X for this post or trim.

4. Skipping the overlay editor on slide 1. Slide 1 is the hook. The overlay text is what stops the scroll. Spend half your review time on slide 1.

5. Not regenerating slides that don't land. The Regenerate button on each slide is one click. Use it. Or use Regenerate-with to swap providers for that slide if a different model handles the aesthetic better.

What to do next

• For the basics of running a carousel-only campaign, carousel campaigns covers strategy.
• For when to swap providers per day, per-day Regenerate-with.
• For platform-specific options beyond carousels (Stories, Reels, first-comments), per-platform composer options.
• For the AI media providers themselves, Magnific (formerly Freepik), explained.