Curmudgeon wrote:
I am going to watch this with interest. Like bsprague, I would like to know more about your workflow, what AI site(s) you use. Creating head shots like that is relatively simple on most AI sites, just the right words in the right order. With AI the devil is in the details, fingers, toes, how objects are held or used. Fingers have improved a lot since I started using AI, toes not so much. Subjects using or holding weapons, ancient or modern, tools of all sorts is still a crap shoot.
I do not use any AI site, I work completely offline... I've downloaded and modded the open source software and also the models.
Everything is open source, no "setup.exe", no support but the forums. If it doesn't work, you fix it.
My system runs on Ubuntu Linux 22.04, an i9-13900 CPU and a Nvidia RTX 4090 GPU (air cooled).
I'm heavily using Stability AI's Stable Diffusion (modded), and ComfyUI (also modded) as a Workflow Manager and UI because of the flexibility. I also use a bunch of python scripts written by myself.
This is the workflow that created the pseudo-photos (as you can see, it's currently running):