Revolutionize Your Content: The AI-Powered Carousel Blueprint

Intro

To create a great carousel with ChatGPT+ you need two prompts and a few minutes.

The first prompt is the long one, but you can easily take the prompt structure out of the first section to modify it as you please, rather than generating five different prompts to pick from. Generating five was a demonstration relatively unrelated to this process, but as I was testing it I found it odd that it chose a topic I had posted about earlier today as the fifth option on the third iteration of testing. Felt like I had to follow through and generate/post that one. I also noticed in testing that generating numerous prompts on a topic gave me options I may not have thought of on my own.

Since posting I've added one sentence to the prompt. "Use size 44 font." If you came from the post this is about, you know why.

For ease of use, and generalization, I removed all custom instructions, and didn't use the system or function roles with the API requests when I used it (not needed for the tutorial).

Initial Prompt

Prompt to generate 5 prompts with difference perspectives

If you want to generate 5 prompts to pick one out use this:

You are a LLM prompt expert. You know exactly how to use the power of NLP to generate the best results by writing prompts for any LLM. We're using ChatGPT and DALL-E 3.

You will be turning a prompt formula like:

“You are [desired expertise].

Write an LinkedIn carousel of text and images for [a targeted persona] to [get a desired result].

Use [needed tone of voice].

Add descriptive visual descriptions of what we can see on each slide.

Add the text of each slide, do not use any emojis, only letters a-Z with punctuation.

Prioritize one unique and uncommon idea as a topic. Ban generic ideas. Ban introduction: jump right into the core of the content.

Add an attention-grabbing first slide that elicits an emotional response.

After all slides are made, please provide a list of prompts for DALL-E 3 image generation, one for each visual description.

Once this is completed, write the VBA code to generate each slide in PowerPoint, leaving the background of the slide blank to be updated with a generated image, but accounting for an image background in the text colors, font, and formatting as well as slide design. Use size 44 font. The PowerPoint presentation should focus on sleek, modern, or even futuristic design depending on the topic. Make it professional, and delete the default title textbox for easier editing.”

To

“You are a makeup artist.

Write a LinkedIn carousel of text and images for young makeup professionals to start their professional journey from growing an online portfolio using Pinterest.

Use a clear, witty and familiar tone of voice.

Add descriptive visual descriptions of what we can see on each slide.

Add the text of each slide, do not use any emojis, only letters a-Z with punctuation.

Prioritize one unique and uncommon idea as a topic. Ban generic ideas. Ban introduction: jump right into the core of the content.

Add an attention-grabbing first slide that elicits an emotional response.

After all slides are made, please provide a list of prompts for DALL-E 3 image generation, one for each visual description.

Once this is completed, write the VBA code to generate each slide in PowerPoint, leaving the background of the slide blank to be updated with a generated image, but accounting for an image background in the text colors, font, and formatting as well as slide design. Use size 44 font.?The PowerPoint presentation should focus on sleek, modern, or even futuristic design depending on the topic. Make it professional, and delete the default title textbox for easier editing.”

Always maintain the structure of the prompt formula and only replace the words in square brackets.

Now, generate a list of 5 prompts without executing them for [insert topic here].

Be sure to generate all 5, in full.

Take a deep breath and work on this problem step-by-step.

**I’ve had some instances of the UI returning 1-3 prompts and saying “repeat for the next [x] prompts” - other times it generates all 5 no problem. I think I noticed a tendency to create fewer the longer a conversation was going on, so I'd try a new chat for each portion of this process. I pretty quickly jumped over to my personal UI to hit the API so I didn’t have to repeat the request while testing.

Just one prompt / write your own:

Alternatively, take the prompt outline from the first portion of the above, fill in as you see fit, and just use that. Here's the outline:

You are [desired expertise].

Write an LinkedIn carousel of text and images for [a targeted persona] to [get a desired result].

Use [needed tone of voice].

Add descriptive visual descriptions of what we can see on each slide.

Add the text of each slide, do not use any emojis, only letters a-Z with punctuation.

Prioritize one unique and uncommon idea as a topic. Ban generic ideas. Ban introduction: jump right into the core of the content.

Add an attention-grabbing first slide that elicits an emotional response.

After all slides are made, please provide a list of prompts for DALL-E 3 image generation, one for each visual description.

Once this is completed, write the VBA code to generate each slide in PowerPoint, leaving the background of the slide blank to be updated with a generated image, but accounting for an image background in the text colors, font, and formatting as well as slide design. Use size 44 font. The PowerPoint presentation should focus on sleek, modern, or even futuristic design depending on the topic. Make it professional, and delete the default title textbox for easier editing.

In either case, make sure to fill in appropriate [placeholders] with what you need.

Now run the prompt and you should have the content of the slides (mostly to keep the AI "focused" for the VBA generation - testing removing this from output), the DALL-E 3 prompts, and the VBA code for PowerPoint.

Image Generation

From here you can copy and paste the list of DALL-E 3 prompts, provided in their own section of the response i.e.

Prompts for DALL-E 3 Image Generation:

"A visionary landscape of a digital city with algorithms as skyscrapers, and beams of light connecting them, showcasing the interconnectivity and potential of Generative AI."

"Illustration of a sculptor, meticulously carving out a statue from a block of digital code, symbolizing the precision of AI in generating and refactoring code."

"A canvas where brush strokes are dynamically forming a modern UI, driven by an AI brain, signifying the automated design process."

"A digital magnifying glass hovering over lines of code, illuminating potential bugs which are represented as glitches, emphasizing the predictive nature of AI in software quality assurance."

"An AI-powered quill writing dynamic documentation on a futuristic parchment, adjusting its content based on the surrounding environment and context."

"A digital interface where an AI avatar communicates with various stakeholder holograms, translating technical jargon into understandable insights, bridging the gap between tech and business."

Into this outline:

Please use these [paste above title + prompts here]

Iteratively prompt yourself after each generation, I would like you to return 4 images for each of the prompts above. If there is an error in generating an image for any one prompt, please skip that image and continue with the remaining prompts and image generation in this initial request.

***EDIT*** The problem with creating walkthroughs for products in beta testing is that they change constantly. You may need to prompt iteratively yourself, as this capability seems to be lost ??

PowerPoint / Carousel Generation

While that’s generating, copy your VBA code from the second prompt. Paste it into a new module of the VBA editor and hit run. (See Google for help on this part if needed)

Now you have a basic slideshow with the text content in place. Go back to where your images should be generated, or just finishing up, select and download which of the 4 you want to use for the slide (If none are what you're looking for, copy the relevant prompt, modify as desired, and generate until you get the image you want - so far I haven’t done this, just on principle for the guide / showing what pure AI with good prompts can produce).

Assuming you downloaded them in order, right click on the first slide, “Format Background…”, in right sidebar select the radio for a picture background, “Insert…” button then go through the slides, the sidebar will stay open to easily update each one.

From there, export the PowerPoint as a PDF, start a new post on LinkedIn, select "More"->"Add a document"->"Choose file", upload your PDF, add a title, and write anything in the post section you'd like - voila! Fastest, beautiful carousel you’ve ever made.

P.S. I highly recommend editing and improving the content - something I didn’t do on principle in the demonstration post.

Huge thanks to @rubenhassid.ai for sharing the prompt that inspired this process, and forms the bulk of the long, initial prompt here. His post can be found here.