I started using Midjourney exactly 2 years ago, mainly driven by curiosity about the technology. The technical side of how visual generative AI works is absolutely amazing to me. That’s why I, like so many others, started early on to experiment and see what these tools can do.
First off, let's look at how it works:
At its core, visual generative AI leverages advanced machine learning techniques to create images.
The AI is trained on a massive collection of images. These images cover various subjects, styles, and scenes, helping the AI learn what different things look like.
The AI uses a special system called Generative Adversarial Networks (GANs), which consist of two parts:
This part tries to create new images based on what it has learned.
This part checks the images created by the generator against real images from the training set and provides feedback.
The generator and discriminator work together in a game-like process. The generator creates images, and the discriminator evaluates them. Over time, the generator gets better at making realistic images based on the feedback it receives.
When you give the AI a prompt (a description of what you want), it uses language processing techniques to understand your request.
The AI combines the information from your prompt with the patterns it learned during training to generate a new image that matches your description.
In the beginning, I used it primarily for two things: 1) creating images for my DnD campaign and 2) testing what it’s capable of.
I’m not that great at drawing myself, and this is exactly the kind of low-stakes personal use AI is great for, in my opinion. I created images of my character and designs for projects within the world of the game, such as flyers for my god and an image for a beverage label (very subtly inspired by Mountain Dew).
I wanted to challenge the AI to see what it can and can’t do. I gave it a couple of prompts that I wanted to repeat with each iteration. This is by no means a scientific approach. For that, I should have timed everything out and tested many more aspects. It was for my own curiosity. But I still found it interesting and fun, which is why I’m sharing it here.
Prompt one: A shark with a pearl necklace drinking tea
(this is the full prompt, I never specified the kind of teacup or settings or anything else)
Prompt two: A seal with a beautiful blonde wig with bangs, sitting on a cliff
.png)
