"Dreams Of Tomorrow", thus named since, quite literally, that's what I fed the attnGAN AI system in the Runway Beta app that created the basic form:
I then took this part, that was only about 256x256 pixels, and passed it through ERNGAN, which specializes in lossless resizing, to get a 1024x1024 version:
I then imported it into GIMP and played around with a LOT of color settings, trying to adjust contrast, chroma values, etc... to give it some colors:
The above also has an engraved transparent version superposed on the less important bits (the middle parts), to emphasize the rest.
I then ran it through Deep Dream Generator:
Looking mighty fine, if I do say so myself !
I then put those 2 results back into the Blender that is GIMP, and combined them using G'MIC-qt :
I then used the Old Photo filter, without Sepia toning, to get this:
And "Voila" !
I hope you liked this little tutorial :)
I highly recommend following Runway on twitter, and to join the Public beta !
They're going for a Credit system, where anyone can run the AI's using store-credit. and given it usually costs 0.50$ for the ENTIRE pipeline (attnGAN x 5-6 to find a good image/saying combination, though once started it's near-instantaneous, ESRGAN x 1) above (each GAN uses ~10 cents of credit, depending which one, and which size input, you choose, and of course how long you leave it running), it's quite a big Bang for your buck !
Just don't forget to stop your models, or that credit will go poof ;)