This new section will showcase my new hobby: muddling along with artificial intelligence artwork. It’s a somewhat controversial technology, but I find it very fun to play with it. And in a way, I feel I’ve the right to use it, as my own POV-Ray images have been used to train those AI models (I found about it through Have I Been Trained?).
I will create several subsections here, grouping images covering the same subject. I don’t claim copyright on any of these images, first because morally I can’t say I’m the author, and second because legally I wouldn’t have a case, I guess. So consider these images to be on the public domain.
Now, technically, I’m using the Gimp Stable Diffusion plugin for generating most of the images. The exception are the images of faces: for these I’m using a combination of Stable Diffusion to generate the source faces at low resolution, and StyleCLIP to generate the final faces using the LQ faces as a source. This allows me to not ever use real persons faces anywhere in the process.
For some images I used the excellent upscaler at Baseten: although it’s really a Photo restoration service based on GFP-GAN, it can be used to upscale images.
Also, I’m not using very complex prompts… just a few words for the description, a few more for the fine-tuning. Most of the time the results are mediocre, but I keep refining and trying the same prompt until I find a seed which gives good results.
Many times I tune the prompt to a general art style, or an artistic period, and some times to a specific artist, but rarely living artists: mostly classic artists whose works are already on the public domain.
Hope you enjoy it!