Hi!

I recently posted some pictures I had created through artificial intelligence on a neural network application called Dall-E mini. Check the pics out here and here.

I wanted to try out Dall-E 2, but the waiting list seems to be too long, and it started to take forever to receive any invitation, so I almost lost all hope to trying out a more efficient AI for creating artificial art.

Now it seems the time has come in the form of Stable Diffusion, an Open Source AI neural network code that has been pre-weighed with pictures found on the Internet. From the github page you can install it on your own computer and use your own craphics card for creating different breathtaking graphical artworks.

There's also a demo page on huggingface.co that uses a free neural network. The same one that was used in the demo version of Dall-E mini.

I started experimenting, and while it's pretty much hit-and-miss, some of the resulting images have been quite interesting.

Here's a couple of the best ones:

I was trying out the rabbit theme I had previously used.

^{A little rabbit wearing a green t-shirt is playing a platformer game on the computer in his room decorated with cyberpunk themed paintings.}

No platformer game though, not sure about the paintings either... (Edit: I realized the platformer game is on the screen on the wall above the laptop, not on the laptop screen! So it actually did draw that as requested.)

^{The little rabbit with a green t-shirt in the room with cyberpunk paintings invests his money into Bitcoin, and gets rich.}

This one was very nice, although much of the description was ignored, again.

^{The little rabbit with a green t-shirt is sad after his bitcoin investments crashed.}

I wonder what happened here.

Foxes

A nobel fox prince in victorian clothing came out looking pretty nice. (It wasn't entirely what I asked though, but close enough.)

^{A fox prince clad in victorian style clothing standing in a room with portraits of other nobel foxes, painted in photorealistic style.}

I asked for it painted in photorealistic style, but it wasn't. Neither were there any paintings around him. Instead the fox prince himself was made into a painting. Maybe the neural network misunderstood my request.

Another fox prince. This time with a wine glass.

^{A fox prince clad in victorian style clothing holding a glass of wine, painted in photorealistic style.}

I'm really impressed in Stable Diffusion, not because it's amazing or anything. From what I've tested I feel like Dall-E mini is a bit better in understanding what you ask, while Stable Diffusion has a lot better quality.

Why rabbits and foxes though?

Well it's partly because we have a lot of them (European Hares, not rabbits, mind you) where we live. Partly because I started my journey into these things by trying with different animals and then just settled with foxes and rabbits.

I also tried creating some pelican paintings in the style of Salvador Dali.

^{A pelican driving a car with mount Fuji in the background, painted in the style of Salvador Dali.}

Just for comparison, here's one created with Dall-E mini (Craiyon) by using the exact same prompt:

Vera wanted me to try out the phrase "a painting of a girl wearing a top hat next to a rose bush":

First the Dall-E created pictures...

And then the Stable Diffusion creation:

Quite a difference in quality I'd say. And it actually looks quite interesting.

Well, I'll be off now, but I guess I will be experimenting more in the coming days. I hope you liked these ones.

See ya!