On August 22nd Stable Diffusion became open for use. I was already seeking for a robust front-end for my favorite early generator, CompVis Latent Diffusion, and SD pushed the envelope further. How could I resist?
My first attempt was a prompt asking for anthropomorphic wasps having a party in a smokey New York apartment, done in the style of David LaChapelle - this was a personal response to wasps having formed a nest in my apartment's wall for a week. There were interesting results and they were glamorous and vibrant, but in error I only saved this single.
After getting the hang of some of it I returned to a early testing staple when I was using Craiyon.
A friend had gained access to Dall-E awhile back and they used this prompt to show me some examples of it. I tried that prompt here with Stable Diffusion - the friend said the results looked more like what they had in mind.
Another early attempt I made when I first used Craiyon or Latent Diffusion was a scene of dogs betting on dominoes in an alleyway (as opposed to playing poker). To remark on this same motif I tried it using an alien, I wasn't sure what to make of these.
In further championing of the wasp nest in my apartment I wanted to try bees in a peculiar style. The last image here has quickly become one of my favorites - see how happy the boy is now that the bees are finally free.
Another favorite from Latent Diffusion was the "Jungle Adventures" magazine spread. With some goading I'm sure Stable Diffusion would realize strong imagery with this motif. At this point I could not use ratios which would have really sealed their beauty.
There were a few other series done but few interesting results. However this last series of an adobe house in the desert left me in some awe - the last image here, I would say, is probably the most realistic I have seen so far in my adventures with generative visuals (though I hardly seek realism).
Stable Diffusion is seriously exciting. In the time it has been made available many tools have popped up, and I'm excited to work with it - especially for larger textural pieces which Latent Diffusion worked well with, or more curiously, video (with folks like @GANWeaving showing great possibility). Midjourney still has appeal for certain imagery but once I have a better production chain established under Stable Diffusion it may become my primary tool for this project.
Stable Diffusion has a free, openly accessible front-end, if you want to try it out yourself. Multimodal.art also has a collection of links for more adventurous explorers.