Which feels a bit sketchy(?) to me, seeing as all models are built on imagery scraped from the net without anyone's permission. It's one reason why I've spent the last month training my own models using my own imagery. If these stock photo sites had any brains, they would also start training models on images in their databases, especially since they already have everything sorted into categories based on keywords (which I'll spend the next year doing, until I can get img2text tools working in recursive batch mode).
Thanks! I have a g-doc where I've been documenting settings and progress here[0] although I just now realized I might not have been using my diffusion model correctly for most of my tests. Iteration 509 of my model I seem to have finally nailed it though! :) I partially "blame" Visions of Chaos since the amazing dev (or devs?) drops updates almost every day with new Machine Learning features, model training was only added recently. I must have reset something on accident.
Also I realize there's a lot of image prep work required, not to mention I have a less than ideal amount of VRAM (3060 Ti w/8GB but no monitors attached i.e. 8GB free) so I have to lower some settings. The source images have to be in 1:1 format (which none of my photos are) so I'm using a script to batch call ImageMagick's 'convert' to add white borders to the top/bottom, which results in my renders also having white borders.
