stable diffusion

I saw some others do guided AI mixed with digital painting, so I wanted to try it myself. Sorry if you hate AI and all that.
.
Some notes about what I did here... I started with an old sketch I did earlier this year. I hastily applied some reddish skin tones and threw it in SD img2img with ControlNet depth filter so I could crank up the creativity but keep the shape and silhouette, then played with the settings until I got something I liked. I'd paint over the parts I didn't like, throw it back in SD img2img, take the output, mask out the parts I wanted to keep, and painted over the parts I didn't like. I saw some others isolate different items but I didn't have any luck with that method. It was much easier to generate the whole image and mask afterwards.
.
I didn't use the stable diffusion plugin in photoshop or anything, I prefer my iPad to my Cintiq, so It was a little difficult to pass around the images between my iPad and my PC. Definitely slowed me down. However, this whole process only took me 3.5 to 4 hours. Not bad.
.
I had a hard time with the feet of course so I had to render the foot myself. Also, the model I was using really wanted to give her ridiculously sized anime eyes. So that was a constant battle. Some other difficult areas were the elf ear, the arm pit, and skin tone overall. I wanted her skin to be red, cause devil succubus or whatever. Not happening with SD apparently. Also it kept removing my blue reflective tones, so what you see in the final was me adding them back in each step. There was a LOT of painting over things.
.
This is definitely not something I plan on doing a lot of. Every AI art, even this one here I did, has this weird AI glaze over it. You can just tell it's AI instantly no matter what you do to it. It was a fun experiment! But I'm going to stick to refining my digital painting skills :D.
.
Tech Info:
Nvidia 2080 w/ 8gb VRAM
Stable Diffusion + Anythingv3 + ControlNet Plugin
614x1024 images generated then masked
Upscaled 2x with R-ESRGAN 4x+