This work was created iteratively using AI to take a rough sketch and transform it into a finished product.
After sketching, I added basic color and shading to get a two tone skin effect. After getting the color and shading about right, I began adding detail using a combination of Local Area Latent Upscaler (available in A111 in the generated list of downloadable extensions) and inpainting to refine different parts of the image. After I'm satisfied, I begin cleaning up the image using a combination of low denoising runs through img2img and/or going into GIMP to touch things up by hand.
The AI is still doing a great amount of work for me, but the composition is mine. Details are there because I wanted them to be. Using the LLuL/inpaint method above, you can mask in new detail or touch up existing detail and not stray from your initial sketch.
I used three mixes to iterate to this result depending on what I needed at that point, with one for detail, another for lighting, and a third for skin texture and facial features. Mixes are not listed because they are not available online.
No prompts with artists or image hosting/portfolio/stock image etc. sources were used to produce this image in whole or part.