Xerew

Like any artist in her gallery's stable who desired to remain as such, Xerew was required to attend all opening night celebrations during which at least one of her works hung on the walls. But instead of schmoozing & selling herself or her paintings (as the gallery would obviously prefer she do), she tended to find herself more interested in determining who amongst the crowd was actually human (and if so, just how human). Over time, her little guessing game became increasingly challenging. Tonight alone, there were at least two people cosplaying as robots (quite convincingly, in full metal regalia, which she imagined they must stay quite fit to even move around in), other people who had modified their own bodies so frequently & to such an extent that the presence of any remaining original parts was debatable, otherwise lifeless humanoid robots that had been sent to the event as physical proxies for their stay-at-home remote human operators, robots that contained within their memory chips uploaded, ongoing versions of their deceased human owners' consciousnesses, & fully autonomous, intelligent robots encased in literal living flesh suits, outwardly (and to a limited depth, inwardly) indistinguishable from human bodies, whom she thought of as "inverse cosplayers." She supposed she should be thankful that amongst all this active self-transformation & elaborate camouflage, not to mention the political turmoil & mechanized police state on constant patrol just outside, that there was still any interest at all in the art on the walls.

[…for more of Xerew's story, visit the official website @ https://inhumantouch.art ...]

ABOUT MY PROCESS

I like using an art gallery or museum space as a setting for my initial AI-generated images, as it adds an intriguing meta-layer to the proceedings, where the AI also generates the various pieces of art being displayed within the main image. This particular image has been worked on for quite some time, on and off, over the course of the past 2 years. I've included the original, failed first attempt, made with DALL-E 2 back in November 2022 (when the engine suddenly started spitting out shit and only shit, IMHO, and I was forced to move on to find other methods to help me continue my pseudo-photoreal process).

Using the character from the DALL-E version as an image prompt in Midjourney, I generated the basis for the girl seen in the final version, although she and her outfit have been inpainted and enhanced repeatedly using Midjourney, Krea, Flux DEV, and Adobe Photoshop generative fill. The crowded gallery that Midjourney produced at first had to be completely rebuilt in Photoshop, to even out the lighting (which tends to get mighty splotchy when you keep zooming out / expanding an image's frame) and the floor texture, and in the end, every other character in the crowd was culled from a separate Midjourney or Flux generation, using cropped details from the working version as image prompts.

Additionally, I took the time to create the final in-world artworks, starting with image prompts based on the more-or-less-vaguely-rendered elements originally positioned within the frames when the gallery was first generated. These used all of the above AI's, as well as MagicQuill and DALL-E 3 (which was needed for the approximation of an anatomically correct heart in the 3rd painting). As always, everything gets composited and further manually refined in Photoshop. The large, regular pattern of the black window frames was drawn by hand (wait, aren't computers supposed to be better than humans at drawing straight lines and complex repeating patterns?), and the rain/fog was added in Photoshop, an easy effect that I've found looks much better when achieved manually, as opposed to leaving it up to an AI.