The short answer: don't screenshot your video, rewrite it. A carousel that works is your video's argument retyped as slides, not a strip of paused frames with the subtitles still burned in.
A ten minute video is one idea unfolding slowly. A carousel is six to ten cards, each holding one idea whole. Getting between them is a writing job first and a design job second, which is why no export button can do it for you.
Why is a carousel the best target for a repurposed YouTube video?
Because a carousel is built for the two things a good video already earns, attention and shares, and because it keeps working after the day you post it.
Instagram has published a plain-language explanation of how Feed ranking works. It lists your own activity first, "Posts you've liked, shared, saved or commented on", and says the five interactions it weighs most heavily in Feed are time spent on a post, comments, likes, shares and profile taps (Instagram Ranking Explained).
Now think about what each format asks of a reader. A single image gives one beat and one chance at a like. A Reel competes with every other Reel on the planet for the same three seconds. A carousel asks for swipes, and swipes take time, which is the one signal on that list a static post has almost no way to produce.
That is reasoning about a mechanism, not a promise about your numbers. Nobody outside Meta knows the weights, and anyone quoting you an exact "carousels get 2x more reach" figure is repeating someone else's sample. But the shape of the format lines up with the signals Instagram says it looks at, and that is a better reason to pick a format than "carousels are trending right now".
The second reason is quieter and probably matters more. A Reel is a broadcast. A carousel that explains something is a document. People save documents, and a saved carousel gets reopened weeks later, which is not something that tends to happen to a 20 second cut of your intro.
A carousel is not a summary of your video. It is the same argument, rewritten for someone who will never press play.
Why does screenshotting your video make a bad carousel?
Because a 4:5 slide keeps only about 45 percent of a 16:9 frame's width, and the part it keeps is rarely the part that mattered.
Run the arithmetic yourself. A 16:9 frame is 1.78 times wider than it is tall. A 4:5 slide is 0.8 times as wide as it is tall. Crop one into the other at full height and you keep 0.8 divided by 1.78, which is 45 percent of the original width. More than half the frame is gone, and you do not get to choose which half unless you crop it deliberately before uploading.
Three more things break at the same time.
Burned-in captions. Subtitles sized for a full screen video shrink to unreadable mush in a feed, and because they are pixels, a typo in them is permanent. You cannot fix a word that has been baked into a JPEG.
Resolution. Pause a 1080p video and you have a 1920 × 1080 frame. Crop that to 4:5 and the most you can salvage is 864 × 1080, which is narrower than the 1080 pixels Instagram wants. Every slide is now an upscale of something that was already compressed twice.
No hook. A frame from minute four means nothing to somebody who has not watched minutes one through three. Slide one is the entire post as far as the feed is concerned, and a random still is the weakest slide one there is.
| Frame grabs | Written slides | |
|---|---|---|
| Shape | 16:9 forced into 4:5, about 45% of the width survives | Built at 4:5 from the start |
| Text | Pixels burned into the image | ✓ Real type you can re-set |
| Fixing a typo | ✗ Re-export the video | ✓ Click the word, retype it |
| Hook on slide one | ✗ Whatever your face was doing | ✓ The sharpest claim you have |
| Reader who never watched | Lost by slide two | Gets the whole argument |
| Time to make | Minutes | Longer, unless something drafts it for you |
What does a good Instagram carousel actually look like?
A good carousel is your video's argument rewritten as a short document: one hook slide, one idea per slide after it, six to ten slides total, the same type and layout throughout, and a final slide that asks for something.
Slide one is the hook. Biggest type in the set, one promise, no logo, no "swipe to learn more" filler taking up the space the promise needs. It is competing with a thumb, not with other carousels.
Slides two onward carry one idea each. One claim, stated in a sentence a reader could argue with, plus a line or two of support. If a slide needs two claims, it is two slides. If it needs a paragraph, it is still a video.
The type is the design. Same margins, same two type sizes, same single accent colour on every slide. Consistency is what makes a set of slides read as one thing rather than seven unrelated posts, and it is also the part that takes the least skill to get right.
The last slide asks. Watch the full video, follow for the next one, comment a word to get the link. One ask, not three.
Every word on every slide should be real, editable text. Not a rendered image of text, not a generated picture that happens to contain letters. Text you can click and retype is text you can fix at 11pm when you spot the typo.
What size should an Instagram carousel be?
Build every slide at 1080 × 1350 pixels, 4:5 portrait. That is the tallest shape a feed post can take, so it claims the most vertical space on a phone screen, and Buffer's size guide describes 1080 × 1350 as the format Instagram recommends for feed posts.
The rule that catches people out is the crop. Meta's own publishing documentation states that "Carousel images are all cropped based on the first image in the carousel", which means slide one silently sets the frame for everything behind it (Instagram Platform content publishing docs). Mix one square in with six portraits and Instagram makes the decision for you, and you cannot adjust that crop after upload.
Two more details from those same docs, worth knowing before you build: a carousel posted through Meta's API holds "up to 10 images, videos, or a combination of the two", and JPEG is the only image format that API accepts. Six to ten slides is the right range anyway, so the cap rarely bites. Export as JPEG unless you are posting manually from a phone.
How to turn a YouTube video into an Instagram carousel, step by step
Five steps: pull the transcript, cut it down to claims, write the hook, put one claim on each slide, then design once and repeat that design across the set.
1. Pull the transcript
YouTube already made one. Open your video, expand the description, and click "Show transcript". Copy the whole thing. Auto-captions are messy and unpunctuated, and that is fine, because you are mining them for sentences, not publishing them.
2. Cut it down to claims
Read the transcript looking only for sentences that survive on their own. A claim is something a reader could disagree with. "Let's talk about pacing" is not a claim. "Most slow edits feel slow because the cut lands after the beat instead of on it" is. Aim for five to eight, and delete everything else without mercy.
This is the step that decides whether the carousel is any good. Ten minutes of video usually contains three real ideas and seven minutes of getting to them.
3. Write the hook from the sharpest claim
Take the claim that would surprise or annoy the most people and put it on slide one. Do not write "5 tips for better edits", because that is a description of the post rather than a reason to read it. Write the thing the tips are actually about.
The test: would a stranger who has never heard of you, mid-scroll, swipe on this? If the honest answer is no, the rest of the slides do not matter.
4. One claim per slide, roughly 12 to 20 words each
Put each surviving claim on its own slide and trim it until it fits comfortably at a size you can read without squinting on a phone. Support it with one short line underneath if it needs one. Resist the urge to add a second thought to a slide that feels empty, because empty is what makes it readable.
5. Design once, then repeat
Pick one margin, two type sizes, and one accent colour, then use those same three on every slide. That is the entire design system you need. Number the slides if you like. Keep the layout identical so the reader's eye lands in the same place every swipe.
The final check: read only the largest line on each slide, in order, skipping everything else. If that alone tells a coherent story, the carousel works. If it does not, the words are wrong, not the design.
Where DesignerOP fits in this
DesignerOP's video-to-carousel maker runs the mechanical half of that process for you. You drop in a video, it works through the transcript, pulls out the claims, writes a hook, and hands back a draft carousel already laid out at 1080 × 1350. You then fix whatever it got wrong.
The part we care most about is what happens after the draft. Every slide opens in a layer editor where the background, the type, and any image are separate, independent layers, and the text is real editable type rather than pixels. Retype a headline, nudge one line, change an accent colour on all seven slides. The draft is a starting point, not a lottery ticket.
That is deliberate. Automation is good at reading ten minutes of transcript and bad at knowing which claim will land with your specific audience. So the maker does the reading and you do the judging. It is the same principle running through the rest of the DesignerOP tools, and we wrote it out in full in how DesignerOP is different.
Being straight with you: DesignerOP is prelaunch. There is a waitlist and nothing you can use today, so treat this as a description of what we are building rather than a review of a shipped product. If your repurposing runs the other direction too, from an Instagram idea back to a video, rebuilding a thumbnail's structure from a reference covers the same reference-first approach for YouTube.
The one thing to take away
Your video is the research. The carousel is a new piece of writing built from it. Once you stop treating repurposing as a cropping problem and start treating it as a rewriting problem, the design part gets easy, because good slides are mostly just one good sentence set large enough to read.