MidJourney — AI Art

Day 24 / 365

I have been working a lot with ChatGPT a lot lately, and other AI models that are text-based. And while it's mind-boggling enough, there is another whole branch of AI that has been blowing up recently, AI Art.

Just like GPT was trained on a huge amount of text available online, and it learned how to “Talk”, there are image-based AI models that were trained with a tonne of images from the Internet, and they learned how to draw.

There are plenty of AI models out there, but the most popular one right now is MidJourney. You ask the AI for any picture, and it will create it for you.

For example, when I asked it to generate a picture of “Pikachu eating dal chawal” this is what it generated —

And it's not as if it's searching from a database of existing images.

It knows what “Pikachu” is.

It knows what “Dal Chawal” is.

And it generates an image based on that. While it looks impressive as it is, this is a free beta version of the product and I am a total beginner at it. Still the results I have been getting are outstanding.

People have already started using this to make characters for their comics, and sprites for their video games. People are even using it to create website designs. It’s shockingly fast as well. The only tough thing is using the right words in your request to the AI.

You can even feed it your own image as a starting point and ask it to use it to generate something new. For example, start with a selfie and see what you would look like as a character on Game of Thrones.

It's shocking how good this is, and you would have to use it yourself to believe it. The best part is that this is not just a gimmick, it has actual use cases. And that’s why I will be exploring this in more detail in the next few days and see if I can come up with a product using it.