aesthetic illusions
patreon
158alt - MaryE W - Cyberpunk
🕑 Added 2024-03-14 14:54:19 +0000 UTCSaw this name and theme combo in the request list so thought I'd make it happen :)
Another of the new Alternative Sets! These are intended for fans of the last set but may prefer a different subject in the images. Keep in mind these are simple face switches, so some of the images may not have swapped nicely, these are intended as bonus sets so they do not undergo the same editing scrutiny.
Check it out in the mega link found in your Welcome Message. This link updates regularly and will also be sent out as a PM monthly. Be sure to click this from a web browser, the apps don't like the welcome message for some reason.
4k versions of all sets available in the Diamond Club Tier!
Note: Patreon has terms of service rules regarding AI images depicting real world people, See here for more. I hope you like what you find!
Comments
Will
I'm pretty transparent about my prompting since there are some proprietary steps going on in the background that make the whole process pretty difficult to replicate, so some quick info: this particular set depending on the pose/position has about 155-200 words used in the prompts including LORAs and textual embeddings, and the scene composition variations (variations which aren't prevalent on this set since the cyberpunk theme LORA was weighted a bit heavier than I'd like so it takes over a lot of the imagery such as dark alley with neon signs in the background) have a wildcard system found in the extension Dynamic Prompts which allow me to save multiple chunks/modules of prompts to a text file that get shuffled into the primary prompt between image generations. this way clothing and background aren't literally the same every picture, and I also don't have to manually add variation between images in the batch. any time an image has been cleaned up a bit, it loses its metadata and can't really be assessed for the prompts used, but every few images you'll be able to drop into stable diffusion's "PNG info" tab or any other meta data reader software and it will bring up the exact prompt used and the particular wildcards used for that image. an example from the VictoriaJ set that would allow meta data assessment would be 158a (16).jpg , but all the 4k images and the alt set swaps have had their meta data scraped off during the processing stage.
Breagan
Awesome!! Would love to see how complexed your prompts are for this type of image.
