Lyssa

Lyssa remains the only @inhumantouch girl to not have her own story blurb, because I used her to demonstrate my general process for making these (see The Steps). These aren’t hard and fast rules: Step 2, for instance, is actually rarely necessary (and only sporadically produces a face better-looking than the original variant), and there is usually some amount of Photoshopping between almost, if not every, step (the human is ALWAYS touching it up, including using it to most often combine the best features from both the original and restored faces), before it gets its own official step toward the end.

Lyssa was one of the first girls completed for @inhumantouch, and thus I was still starry-eyed with wonder at the very existence and accessibility of AI-text-to-image technology, remaining relatively hands-off when dealing with DALL-E's output, whereas soon enough I began to regularly combine elements from disparate generations with Photoshop, sometimes resorting to Midjourney or Stable Diffusion or even Adobe Stock to generate or acquire certain elements I wanted present (once the apparent theme was more solidified), and learned what cases required me to reduce or enlarge the central part of the image to get the best results when outpainting the next level.

With almost 60 of these girls completed now, DALL-E’s method/madness still remains largely a mystery to me. It doesn’t have a full slate of command-line parameters one can tweak like the other AIs do, and I’m not one to enter paragraph-long prompts in any case, but that seems to leave it up to chance whether it tends to be cooperative/inventive/contributing to our “creative partnership,” or seems just plain disinterested in what I’m trying to make, if not actively sabotaging my attempt. For something that is highly unlikely to have gained its own consciousness yet, it still behaves as if it’s got quite the personality.