How to make image carousels for your app that people save

What a carousel is for, and what it is not

A carousel is the post people keep. A video is the post people are shown. They are not substitutes, and picking the wrong one for a topic is the most expensive mistake in this format.

• Carousels buy dwell time and sends. Instagram's head Adam Mosseri has said carousels reach more people than single images, because more media means more chances to interact. A carousel nobody swipes also gets a second chance: the feed auto advances it to slide 2.

• Engagement is much higher, reach is about the same. TikTok's own comparison of photo posts against video lands around twice the likes and roughly three times the comments and shares, on a few percent more reach. More interaction per viewer, not more viewers. These are vendor-reported figures, so treat them as directional.

• The tail is longer. A photo post keeps collecting views into weeks two, three and four, where a video has usually spent itself in week one. That changes what a bad first day means.

• The format loses when the content is personality. Education, lists, comparisons and checklists win as decks. A joke does not want to be swiped, and a reaction does not survive being frozen.

• TikTok photo mode is a quieter room. It is a less competitive discovery surface than the video feed, which cuts both ways: less to beat, fewer people scrolling it. Photo posts are also not eligible for TikTok's Creator Rewards, which only matters if you were counting on payouts.

So measure a deck on saves, sends and profile visits. Views is the wrong number for this format, and it is the number the app shows you first.

The shape of a deck

Five to ten slides, and the three parts do genuinely different jobs. Instagram caps a carousel at ten and drop-off sharpens after about seven, so ten is a ceiling, not a goal.

1. 1

   The cover. One hook, nothing else. No logo, no handle, no subtitle competing with it, no pitch.

2. 2

   The body, three to seven slides. One point each, and no slide may depend on the one before it unless the pattern is a sequence. The count comes from the promise on the cover, so if the topic honestly supports four points, make four and do not pad to seven.

3. 3

   The closing slide. One ask, tied to the thing you just taught. This is the only slide that names your app, and the only place a plug belongs.

Never number the body slides. A visible “5.” right before the last slide is a completion signal: readers stop there and never reach the ask. The count belongs on the cover, where it earns the swipe. Both platforms show position anyway, with a dot counter on Instagram and a progress bar on TikTok.

The cover is the whole post

Most people who see your carousel see the cover and nothing else. Write it as if it is the only slide, because for most of your audience it is.

• Open the reader's problem, not the topic. “5 budgeting tips for freelancers” is a table of contents. “Your budget survives the month. Your money leaves on Tuesdays” is a hook. One labels the deck, the other opens a loop.

• Keep it to about twelve words. Past that it wraps to a fourth line, the type has to step down to fit, and it stops being readable at feed size. Three comfortable lines is the real ceiling.

• Digits earn the swipe. “3 reasons” tells a reader how long this will take, which is an honest reason to start. Numbers are not clickbait. “You won't believe”, “secret” and “hack” are.

• Promise the payoff, keep the content. If the cover already gives the answer, there is nothing to swipe for. The payoff lives on slide 3.

• Never sell on slide 1. A pitch on the cover is read as an ad and skipped as one, and it costs you the whole deck rather than one slide.

Write five covers for one deck and post the one you would stop on in somebody else's feed. It is the cheapest part of a deck to change and the only part most people will ever read.

One subject per slide

A slide illustrating three things illustrates nothing. Write the sentence first, then list the objects that sentence actually involves, and let that count decide the layout. The count is not a detail: it is the layout.

• One object: give it the frame. One thing drawn large in the bottom quarter, or cropped hard by an edge so only a corner of it is inside the frame. The scale is the whole effect.

• Two objects: pair them. Either a loose diagonal, one in the bottom-left and one in the top-right, or both sitting together low in the frame. The diagonal reads as a contrast, the low pair reads as a set. Pick the one your sentence means.

• Three objects: three corners, fourth left empty. Leave one corner bare on purpose. Four objects in four corners is a pattern; three and a gap is a composition, and the gap is where a long headline grows. A row along the bottom edge is the other honest option.

• Four or more: a border or a scatter. Ring the edges with them and leave a break at the top center, or drop them small and far apart near the corners. Both keep the middle open by construction rather than by luck.

• Never pad the list. A fourth object does not decorate the page, it turns it into a different page. If the sentence involves one thing, the slide gets one thing.

• Six to nine objects is the readable ceiling. Anything busier, plus a headline, plus a supporting line, is clutter at the size people actually read it.

The test is one question: what is this slide about? If the honest answer has an “and” in it, it is two slides.

Keep the center of the frame clear

Reserve the middle 60% of every slide for type and put the art at the edges, in the corners, or in the bottom quarter. This single rule separates a deck that looks designed from a deck that looks assembled.

• Type cannot rescue itself over art. A bolder weight, a drop shadow, a dark scrim: every one of those is a bet against an object edge you did not plan for, and every one of them looks like a patch. Move the art instead.

• A graze is the tell. The most common way a competent deck reads as amateur is one line of type just touching the edge of one object. Nobody can say what is wrong with it. It simply looks unfinished, usually on the slide that mattered.

• Check it at thumbnail size, not editor size. Shrink the slide to the width of a feed grid cell and look at it on a phone. If the headline is not readable there, it is not readable, and no amount of zoomed-in polish fixes that.

• The bottom and the right belong to the platform. Instagram sits the caption under the frame, but TikTok draws a photo post full screen with its own caption, handle and buttons over your art. The same files go to both, so keep type out of the bottom eighth and off the right edge or TikTok covers it.

• Contrast is a floor, not a preference. Pick a text tone that clears your surface tone by a real margin and hold it across the deck. If you have to squint on one slide, the whole deck reads as low effort.

Words on a body slide

A body slide holds a headline and one line. That is the entire budget, and it is smaller than it feels while you are writing.

• Headline: six words or fewer. It names the point. If it needs a seventh word, it is usually two points wearing one headline.

• Line: one sentence, eighteen words or fewer. Not a paragraph. A paragraph on a slide is a slide nobody finishes and a screenshot nobody takes.

• Separate the two by size and weight before color. Posters get hierarchy from scale, app interfaces get it from faded grey. You are making a poster. Roughly a two-to-one size gap and a real weight step does the job on its own.

• Never use transparency for the secondary line. Ink at 60% is legible on some slides and mud on others, and you will not notice which until it is posted. Use a solid tone mixed toward the surface, so every slide behaves the same.

• Break lines by balance. Two lines of similar width, not one long line and a stub, and never one word alone on a line. That orphan is the second most common amateur tell after the graze.

• One legible face for the whole deck. Handwriting and grunge faces lose at feed size. Let the illustration carry the personality and let the type just be readable.

One style per deck, ideally per brand

A follower should recognize your post before they read a word of it. That only happens if the look stops moving.

• Pick one illustration language and hold it. Soft 3D clay, layered paper cutout, flat editorial line work, watercolor, bold graphic poster, dark studio photography, pixel art. Any of them works for an app. Two of them inside one deck does not.

• Two accent colors and one surface tone. The same three values on every slide of every deck. That, more than the drawing style, is the thread a follower actually recognizes in a feed.

• Never let an image model draw letters. Any object that carries writing, a book cover, a sign, a labeled jar, smuggles broken lettering into a slide you are about to set real type on. Keep them out of the art and set every word yourself.

• No hands, no faces in generated art. A generated hand is still the single most recognizable AI tell, and one of them undoes the craft of everything around it.

• Do not repeat a layout on consecutive slides. Non-adjacent repeats are invisible to a reader. Back to back, the deck reads as one template with the words swapped, which is exactly what it is.

Patterns that work for an app

Pick the shape before the topic. Pick the topic first and everything comes out as a tips list, which is the shape everyone already posts.

• Mistakes. Something the reader is probably doing wrong, plus the correction, one per slide. State it plainly and hand over the fix; never scold. For a budgeting app: “Your budget survives the month. Your money leaves on Tuesdays.”

• Myths. A belief the field repeats constantly, and what is actually true. Correct toward the mainstream position, never a contrarian take you cannot back. For a sleep tracker: “Catching up at the weekend does not undo the week.”

• Steps. A sequence where the order is the content, so a reader who follows the slides top to bottom is finished at the last one. For a language app: “Your first 100 words, in the order that makes them stick.”

• Checklist. What to check before doing something, phrased so a reader can answer yes or no while standing in front of it. This is the shape people save most. For a photo cleanup app: “Check these 5 things before you delete 10,000 photos.”

• Comparison. The same two options on every slide, judged on a different dimension each time, and neither side becomes the villain. For a habit tracker: “Streaks or weekly targets: which one survives a bad week.”

• Red flags. Warning signs to notice in something else, a product, an offer, somebody's advice, and every one has to be observable rather than a feeling. Mistakes are what the reader did; red flags are what the reader should spot. For an invoicing app: “4 things in a first client email that predict a late payment.”

• Numbers. One figure that reframes the problem, unpacked across the body slides. Use a widely published figure, or your own aggregate data (“we looked at 40,000 grocery lists”), and skip this shape entirely rather than reach for a number you would be guessing at.

• Tips. A flat list of independent pointers where the order does not matter. It works, it is the easiest to write, and it is the one everybody posts, which is why it should not be your first three decks.

Two pairs get confused constantly, and one question separates each. Tips and steps look identical until you ask whether the order matters. Mistakes and red flags look identical until you ask whose fault it is.

Where carousels actually publish

This is the planning fact founders get wrong most often, and it decides your week if you are also making video.

• Instagram: a native carousel, up to 10 slides. Export at 1080 by 1350, which is Instagram's tallest feed aspect, and give every slide the same dimensions. The feed crops the rest of the deck to match slide 1, so one odd slide reframes all of them.

• TikTok: a photo post, up to 35 images. TikTok takes the same files. Its photo documentation specifies format, 1080p and 20MB per image and never mentions an aspect ratio, so hand it the 4:5 set and let it fit them itself.

• TikTok's title field is 90 characters and it rejects a longer one. It does not trim for you. A 91-character title fails the post, which is an expensive way to lose a deck you spent an hour on.

• Facebook: a multi-image feed post, Pages only, up to 10. The per-image limit is 4MB, so export JPEG rather than PNG. A clean 1080 by 1350 PNG can clear 4MB on its own.

• YouTube takes video only. A YouTube post requires exactly one video file. There is no image post to plan for, so carousels simply skip it. If your calendar assumes every post goes everywhere, this is the one that breaks it.

One asymmetry worth knowing before you design anything: text you bake into an image is not searchable. TikTok indexes text added with its own text tool, not your pixels, so the words you want found have to live in the title and the description.

Write the title separately from the hook

TikTok prints a title line directly above your deck and Instagram gives you a caption. Neither one wants your hook again.

• The hook opens a loop. The title closes one. The cover is already on screen under the TikTok title, so repeating it prints the same sentence twice and spends your only other text slot on nothing.

• Write the title as the plain label the hook is forbidden to be. It sits next to a thumbnail and it gets searched, so use the words somebody would type: “3 reasons your toddler skips dinner”, not a curiosity gap.

• The caption is a third job again. Two short lines. Skip the hashtag wall: Mosseri has said hashtags help search and sorting, not reach.

• Put the invitation in the first comment. One warm sentence asking for saves and questions keeps the caption clean and gives people something easy to reply to, which is the signal this format is actually good at.

Reusing images: the real risk and the folklore

Most of what founders worry about here is the wrong thing, and the thing they do not worry about got stricter this year.

• Reposting your own work is fine. The originality rules target reposting other people's content, not yours. Nothing stops you re-cutting a deck you made or running the same topic again in a new style.

• Reposting other people's images is the real risk. Instagram extended its originality system to photos and carousels on April 30, 2026. Accounts that repeatedly repost others' content, in their own wording ten or more times in thirty days, lose recommendation surfaces like Explore and the non-follower feed.

• Crops and filters do not count as making it yours. Instagram is explicit that superficial edits do not qualify and material changes do: your own text, real creative edits, a voiceover. Text laid over somebody else's photo sits exactly on that line.

• Pulling slide art off Pinterest is the common version of this. It is also the version where you do not own the image, which is a separate problem that no reach argument will save you from.

• FOLKLORE: “TikTok flags images over 70% similarity and costs you 50 to 80% reach for three to six months.” Those numbers trace back to SEO content farms, not to any platform. No platform has published a similarity threshold, a reach penalty percentage, or a recovery window. Ignore any guide that hands you one.

• Generated art sidesteps the duplicate question. New pixels, and they are yours. TikTok and Meta both describe their AI labels as disclosure, not a distribution signal, so labeling a deck is not a reach decision. The cost moves to craft: people scroll past art that looks obviously generated.

Why most founders never ship one

Worth saying plainly, because nobody selling you a template will: a good deck is more work than a good video. That is the actual reason your competitors post one carousel and stop.

• A reel is one hook and one clip. A deck is nine decisions. A cover, a point per slide, a picture per slide that has to stay out of the type, a closing ask, a title, a caption. Any one of them can be the thing that ruins it, and only the cover gets a second look.

• Your first one will take an hour or two. Budget it. Expect the layout to feel wrong until about slide 4, which is where the style finally settles and the earlier slides need redoing.

• The template trap is the failure mode. The moment you solve one layout it is tempting to reuse it on all seven slides. That is what makes a deck read as filler even when the writing is good.

• Batch the writing, not the design. Write five covers in one sitting and design them across the week. The writing is where a deck is won, and it is the part that does not need any tools open.

This is the part Vubio automates: it writes the deck, illustrates every slide in one style picked for the brand, and posts it to Instagram and TikTok on your schedule. Either way, the rules above are the same rules.

Questions founders ask

• How many slides should a carousel have? Five to ten. One cover, three to seven body slides, one closing slide. Instagram caps a carousel at ten, and reader drop-off sharpens after about seven, so treat ten as a ceiling rather than a target.

• Do carousels get less reach than video? Roughly the same reach, with far more interaction per viewer. TikTok's own comparison of photo posts against video puts likes around twice as high and comments and shares around three times, on about the same reach, and photo posts keep collecting views for weeks longer. Those figures are vendor reported, so treat them as a shape rather than a promise.

• Can I post the same carousel to Instagram and TikTok? Yes. Export one set at 1080 by 1350 and send the same files to both. Instagram takes up to 10 slides at that aspect with no crop, and TikTok photo posts have no documented aspect requirement and resize what they are given. Write a separate title for TikTok, under 90 characters, because it prints one directly above your deck.

• Do AI-generated carousel images hurt my reach? No. TikTok and Meta both describe their AI labels as a disclosure mechanism, not a distribution signal. What platforms demote is reposted content that is not yours, which generated art avoids by construction. The real cost is a viewer scrolling past art that looks obviously generated, which is a craft problem rather than an algorithm one.

• Can I post carousels to YouTube? No. A YouTube post requires exactly one video file, so there is no image carousel to publish there. Instagram, TikTok and Facebook Pages are the three surfaces that take one.

• How often should I post carousels? One or two a week alongside video is a realistic rhythm for a new account. Carousels are also the format to drop first in a busy week, because a deck takes longer to make than a reel and a silent week costs more than a missing format.