I saw a post about an AI package called MidJourney on twitter which peaked my interest. Midjourney is an interesting artificial intelligence model which allows you to feed it a phrase and it will generate pictures / drawings.
For those who are not aware of these types of artificial intellegence models, they basically learn a specific aesthetic and content by analysing hundreds, thousands and hundreds of thousands of images to learn specific styles, and content within an image, this process can be considered as ‘machine learning' as the AI model is given lots of types of art and it learns to udnerstand what is similar in each of these types or styles of art before it can then have a go at reproducing based on prompts it has been given.
It should be noted that AI generated art is an area of great contraversy, and I will not get into that area fully other than to say that there are concerns around copyright, but also creativity (amongst other things) as the AI is simply learning a style and just reproducing a range of elements in that style, but that is a whole post for another day.
I joined the BETA (as it is not yet open fully to the public) to have a look at how the AI model worked, in the current BETA phase to ensure fairness it generates four different images based on your suggestion rather than generating hundreds, but to be honest that is enough to be able ot get an idea of how it works and how well it works.
To test the AI Model, I gave it the phrase “A man and woman laying on the beach looking into the sky with stars and planets”. The below image is what was generated:
I wanted to give it some different styles to see what else it could come up with, so for my second prompt I gave it the phrase “An eagle flying across a river in a renaissance style”, the model came up with the below images:
For the very last test I got my wife to come up with a phrase, she had the phrase “A tiger on the beach at night in the style of Surrealism” which generated the AI generated artwork below:
When you consider what the model is able to do, it is fairly remarkable (concerns aside) that machine learning has been able to learn the different styles of art, but also content within the art itself to be able to place things like an eagle, or stars, or a beach but also in a specific style. If you are interested in finding out more you can do so through the Midjourney website.
Comments are closed.
PO Box 113, Beaconsfield Upper, Victoria, Australia, 3809.