This is a continuation of a thread started by @Num7 on the archived boards.
AI generated images are a great way to create new and interesting visuals for users. From abstract art to realistic renderings, AI image generation can produce results that are often unpredictable and visually stunning. It is no surprise why this form of artificial intelligence has become increasingly popular among designers, artists, and viewers alike.
AI image generation utilizes deep learning algorithms to create photos that do not exist in the real world. By leveraging large datasets of images and details, AI image generators can learn how to combine elements from those sets into a new visual. This means that AI images can often look more realistic than photographs taken by human photographers because they will contain combinations that no one person could capture in real life.
I’m partial to MidJourney because it gives me the most interesting results out of all the ones I’ve tried, and I like that the interface is attached to Discord. There’s plenty of others though, like Dall-E Mini and Dreamscope. If you know of any good ones feel free to link them here.
I’ll use this thread to share some of the interesting images I’m able to generate, and you’re welcome to share some of yours too :). Let’s see what kind of cool AI generated art we can make!’
@Cosmo - I did a variation of the original image, found here, with a prompt of my own. I didn’t like how it looked, so I did a variation of that variation, and I liked that a little better. The two problems I have with the finished product is:
The boy’s left hand appears to have an extra finger.
The boy’s left foot has a couple of toes fused together.
Is there something wrong with the tool, or am I doing something wrong?
I don’t think there’s anything wrong, you just have to keep generating until you get one with proper proportions and keep going. If you make an image from that base image 100 times, you’ll get 100 different compositions. The nice thing about MJ though is you can just keep slamming that regenerate button until it makes something you’re happy with.
When I made the image for the homepage of TTI, I must have gone through 200 different regenerations and variations of “wizard’s tower in a void” type prompts. It’s an art and you’ll not get something exactly perfect within the first few tries unless you’re lucky.
For what it’s worth though, I never would have noticed the kid’s hands being too big. We’re always our own worst critics and notice stuff other people will never pick up on.
The same prompt, but each set of images is completely different. Then, one one of those you could roll variations and adjust the prompt to try and edit it a little.
There’s also some great tools to try and get styles and things into the mix:
Much potential. I think people are only scratching the surface.
With the “birds inside a robot eating fish” prompt, were you able to get the result you desired? And can you actually edit a prompt you already posted?
You can sorta see I hit the “V3” button in the background to make a variation of image 3, then changed my prompt to cow and 1800s photo style. That gave me this:
Hmm, it took me many tries to generate this one. It’s not perfect, but at least it passes. Maybe images of certain things like arrangement of letters and certain national flags are not Midjourney’s forte?
Oh yes, AI is an absolute blast. I’ve been head-down working on some things making use of this and ChatGPT, and the results are scary good. I’m aiming to share some more this weekend.
