This is a new piece of art I created for the Art Explosion contest theme "Surrealism". Make sure to visit the contest post and give contest host @juliakponsford some love - she is awesome! I describe the process I used for this piece of art below.
Dreaming of the Dreamer - Carl Nash
Here are a few closeups of sections:
The Process
I used an interesting process to make this piece, so I wrote a post about it :)
I took some macro photography of a big mass of bubbles in my yard yesterday figuring I could make something interesting out of the cool shapes and swirly colors. This piece of art started life as this cropped section of a photo of bubbles:
I ran this photo through a neural network that I trained in two of my own styles of art using the Deep Dream Generator
First I trained the neural network using this piece of my own art:
Shoggoth in the F ungus Forest - Carl Nash
Here is the result of the neural network generating an image from the bubble photo in the style of my shoggoth drawing - I used this as a semi-transparent layer in the final piece:
Bubbles in Shoggoth Style
Next I trained the neural network in the style of my (profile pic and background image) eyeball alien drawing and ran the first result back through. Here is the eyeball alien I used to train the neural network:
Eyeball Alien - Carl Nash
Here is the result of uploading the Bubbles in Shoggoth Style image and generating a new image from it in the style of eyeball alien:
Shoggoth Bubbles in Eyeball Alien Style
Finally I ran the resulting Shoggoth Bubbles in Eyeball Alien Style back on itself using the standard deep dream settings several times until the neural network had created some interesting things:
Bubble Dream
I took the resulting Bubble Dream image into Photoshop, put the first image I generated from the neural network (Bubbles in Shoggoth Style) as a 50% transparent layer above it, and added another layer on top to paint color on top to accentuate things that I saw in the image. Basically I started running the image through my own neural network. This was a really fun exercise because I had trained the AI in my own style in the first place so it was easy for me to make sense of what was happening - even though it is some trippy nonsense.
Here is just the layer of color that I added in Photoshop - it is kind of strange to see it without the layers underneath it but I think it gives you a better sense of what I altered:
Hope you enjoy the piece and the process!
Cheers - Carl