ABARA Style Manga Lora [Tsutomu Nihei] Z-Image

[testing, will add details here later, or unpublish if not up to scratch. Easier to test gens quickly here]

***NOTE - I'm reaching out to the dude behind SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts | Patreon commissioning him to help out ⭐

Ok, I don't have many friends in the lora/model making arena to 📞 to help with troubleshooting, so I have to figure out where/if I've gone wrong on tagging the dataset, or if I would have been better off using fewer images perhaps. Or if it's a Z-Image natural language preference issue.

So not worthwhile yet to download apologies (unless you can get it to work somehow in which case god bless 😂)

[CURRENT ISSUES]
Prompt: " A raw, high-contrast black and white manga panel in Tsutomu Nihei's Blame! and ABARA style, ultra-detailed 8k scan with heavy scratched texture, dense crosshatching, ink splatters, floating droplets and film grain. Extreme close-up side profile of a terrifying silicon creature: completely smooth, featureless white helmet-visor head — pure Slenderman blankness, absolutely no eyes or mouth on the helmet — yet below it erupts a massive jagged mandible split wide in silent snarl, rows of razor-sharp teeth gleaming against black shadows. Above the helmet explodes a feral corona of splintered bone ridges and shredded membrane flanges unfurling like a tattered executioner's hood torn by hellish winds. Thorax is a cantilevered cage of hyper-elongated ribs covered in cross-hatched decay veins; long sinewy arms end in oversized taloned claws splayed hungrily, fingers curling through airborne grit and disintegrating flesh particles. Lower body fully dissolves into roiling black particulate chaos and ink void — no legs, swallowed by negative space. Creature floats alone in infinite featureless dark abyss, carved out only by a single dramatic rim light that traces razor-thin white highlights along silhouette edges. Maximum dread, unsettling horror, masterpiece detail on every bone shard, membrane tear and ink scratch"

⬆️⬆️ ⬆️ This style of prompt in Seedance or Nanobanana Pro delivers what I want more or less, but not with Z-image and my lora.
In fact, minus the lora and just rawdogging a generation, Z-Image on its own doesn't even make any ABARA style images that I like.

I got cocky with the H.R Giger model coming out close to OK with minimal effort, so back to the drawing board for this one.

Not worth downloading (Yet🙏)

Any tips you might have for improving quality would be very appreciated