How to create a carousel with Gemini
GEMINI TIPS
Phillip Twyford
If you make video content, you've probably got a backlog of it sitting untouched. You filmed it, you posted it to YouTube, and that was the end of its life. I've started using Gemini to pull a second piece of content out of every video I make, in the form of a ready-to-post Instagram carousel. Here's exactly how I do it.
Why a video summary alone isn't enough
The obvious move is to paste your YouTube link into an AI chat tool and ask for a summary. That gets you a summary. It doesn't get you a carousel.
The gap is structure. A carousel needs several distinct cards, each with a job to do: a title, an overview, a handful of key points, and a takeaway. A flat summary doesn't give an image generator anything to work with. If you skip straight from raw video to "make me some images," you'll get generic, unusable results.
The fix is to treat this as two separate jobs, in two separate chats.
Step one: get a structured breakdown of your video
Open Gemini and paste in your YouTube link. Ask it directly for a detailed summary broken down into seven key areas. I do this for every long-form video I publish, because it forces the summary into a shape that maps onto a carousel before I've even mentioned images.
Once Gemini comes back with the breakdown, read it properly. Don't skip this. AI tools miss nuance and sometimes invent detail that sounds plausible but isn't what you actually said. Check the seven points against what's in your video. If something's off, correct it before you move on.
Step two: start a fresh chat and brief the carousel
This is the step most people miss. Open a new chat. Paste in the checked summary and label it clearly as a video overview.
Then give Gemini a specific instruction. Tell it you want an Instagram carousel, sized 4:5, made up of seven separate images. Tell it the visual style you want, for example a modern digital illustration with a clean, friendly look. Then go card by card. For each image, give it the title, the purpose of that card, and what you want it to actually show.
The specificity is what makes this work. A vague instruction gets you a vague carousel. A card-by-card brief gets you seven images that follow a real structure: a title card, an overview card, individual cards for each key point, and a closing card with the main takeaway.
What you get back
Gemini generates all seven images in one pass, matching the style and content you specified for each card. You can download them and build your carousel in whatever tool you normally use.
The result reads as a proper carousel, not seven disconnected AI images stitched together. That's because every image had a specific job before it was ever generated.
Where this fits into your content process
This works for any video you've already published. You don't need to film anything new. If you've got a backlog of YouTube videos, each one is a carousel you haven't made yet.
It also works well alongside a LinkedIn post drawn from the same video. One recording, checked once for accuracy, gives you three formats: the original video, a written post, and a visual carousel.
What to do next
Pick one video from your back catalogue this week. Run it through Gemini for the seven-point breakdown, check it against the video, then brief the carousel in a fresh chat.
If you get stuck on the image briefing step, start simple: title, purpose, and one line on what the image should show for each card. You can refine the style once you see the first result.
If you want to learn more about the incredible and free Gemini Notebook, click here.
FAQ
Q1: Can I really make an Instagram carousel from a YouTube video I already published?
A: Yes. Paste the video link into Gemini, ask for a seven-point breakdown, check it against the video for accuracy, then brief a fresh chat to generate the carousel images. No new filming needed, and every past video is a carousel you haven't made yet.
Q2: Why isn't a basic AI summary enough to generate a carousel?
A: A summary is a flat block of text with no structure. A carousel needs seven distinct cards, each doing a specific job (title, overview, key points, takeaway). Without that structure, an image generator produces generic, disconnected results instead of a proper carousel.
Q3: Why do I need two separate chats instead of one?
A: The first chat turns your video into a checked, structured seven-point breakdown. The second chat uses that breakdown to brief the actual images. Combining both steps in one chat skips the structuring work, which is what makes the difference between usable and generic output.
Q4: How specific do I need to be when briefing Gemini for the images?
A: Very specific. Give the format (4:5, seven images), the visual style, and then go card by card with a title, purpose, and what each image should show. Vague instructions get vague results. A detailed, card-by-card brief is what produces a coherent set.
Q5: Should I trust Gemini's summary of my video without checking it?
A: No. AI tools can miss nuance or invent details that sound plausible but weren't actually said. Read the seven-point breakdown properly and correct anything inaccurate before using it to brief the carousel. This check happens once, before either downstream step.