Generate the Perfect Slides - Part One

I am writing this from a train 🚊 under the sea 🌊 on my way to London 🇬🇧 for Big Data London and to present at the Tableau User Group.

Quite cool, right?

I’ve always enjoyed making presentations. I like designing slides and speaking in front of people. However, I’ve always struggled to find the right pictures or logos that perfectly represented what I had in mind… until recently!

In my next two publications, I would like to share some helpful tips about how to create slides using Midjourney. It's important to note that this is not an AI Newsletter, and I'm not an expert on prompts.

However, I have been using Midjourney for my recent presentations and have developed a process that helps maintain a consistent style and improve the quality of images with an excellent free tool. This process has dramatically improved my presentation-making experience, and if you sometimes struggle with creating presentations, I have some valuable tips that can help you, too.

This tutorial is quite long, so I’ll separate it into two parts. The second part will arrive next Thursday!

Generate the first picture.

The first part is defining a style and generating the first image. This is, for me, the most challenging part. I can spend hours trying different words and styles until I reach something that I enjoy.

The title of my talk is: “Building Bridges, One Path to Success with Tableau,” and I already have a style in mind; I want the pictures to be mostly outlines and hand-drawn. The theme of the User Group is Sherlock Holmes, so I also want something resembling old paper as the background.

This is the first result (you’ll find the prompt below each result):

I don’t know for you, but I am not a fan of this result (and I also forgot to put —ar 16:9 to have a ratio for slides).

When that happens, I either regenerate four new images with the same prompt or tweak it a bit. There’s a fundamental randomness in image generation, so it sometimes takes a few tries to find the right style.

Let’s try again with a slightly different prompt:

The first one! That’s the style I have in mind. Let’s use this image as our starting style for all future image generations.

Now that we have one picture and a style, generating the other picture will be easier and fun! Let me show you how to keep a coherent style.

Planting the right seed

I’m combining two techniques to generate new images in the same style.

The first technique is straightforward and obvious: keep the same style-defining word in each prompt. For every new image, I’ll keep it “as a minimalist, outlined drawing” at the end. Simple.

The second technique is more advanced and is where the real magic happens. Every picture in Midjourney is generated using a random seed. That seed is like a starting point for your image that will orient it in one direction.

You can get the seed of one image and use it to generate new ones. This will ensure that the overall style or composition stays the same across different generations. Here’s Midjounrey’s official explanation:

Here’s how you can find the seed of one picture and use it to generate new ones in the same style.

First, you need to have an upscaled version of the image. You get this by clicking on the “U” option of the image you want to upscale. In my case, it’s “U1” (the “V” option will create variants of that image).

Once Midjounrey sends you the upscaled version, you can simply add an “envelope” reaction to the post. The Midjourney bot will answer you with the image, prompt, job ID, and the precious seed:

That’s it; you can now use the same seed to generate new images. All my following prompts ended with “as a minimalist, outlined drawing --seed 1453635179 --ar 16:9” and, as you can see in the next two examples, it worked pretty well!

As you can see, the general style and color are similar between the different images.

I liked the image in the bottom right, but there’s something not coherent: people are walking in the “water” below the bridge… We need to fix that!

Sadly, my train arrives in London in a few minutes, and I have an important presentation tomorrow, so I’ll need to keep you on your toes and keep the second part for next week!

Next Thursday, I’ll show you how to fix mistakes in generated images and upscale your pictures to have crisp lines and no artifacts when putting them on a big screen!

That’s it for this week!

You can reach us easily over there:
‣ Website: https://www.ladataviz.com
‣ Twitter/X : https://twitter.com/ladataviz
‣ Youtube: https://www.youtube.com/@ladataviz
‣ LinkedIn: https://www.linkedin.com/in/ladataviz/
‣ All other links: https://linktr.ee/ladataviz

If you have any suggestions, feedback is appreciated!