<h2 id="successor-is-in-development!">Successor is in development!</h2>Join the <a target="_blank" rel="ugc" href="https://discord.gg/8Kuywkmtve">Cult of Gremmy discord!</a> ==--== illustrious version notes==--== ——— Avarice:In a world smothered by rain and neon rot, Avarice bleeds style and wrath in equal measure. The realm is one of iron and ember, where arcane circuitry coils beneath obsidian leathers and each step is etched in radiant crimson.This LoRA evokes a singular vision: tall silhouettes wrapped in long, ritual-stitched coats; eyes like fire behind Urban hellscapes frame streets lined with flickering glyphs, steam curling around platform boots that sound like judgment on the pavement.It is the aesthetic of power dressed in precision: cyber-occult. Urban-deviant. Beautifully damned. Every image sears with controlled violence, red runes like warning flares around the bodies of men and women who do not flinch.Hair like shadow. Skin like prophecy.---For reference, I do not gen locally. Personal settings:CFG Scale: 1-3.5 (it does WAY better at low CFGs)ReccomendedLora strength/weight for version 1.0: .1-.7 (slightly overtagged the dataset for 1.0 sorry)Sampler: Heun (softens image, gives very painterly and unique style), DPM/DPM 2A, DDIMSteps: 40 for Heun/DDIM (can also go very very low steps and hi res fix for a sharper background), 18-27 for DPM, DPM2A.

Avarice

I'm happy to announce V1.0 (originally trained as 0.4) of Avarice; A 1.1 may come in the following weeks as more data comes through in order to slightly tame the strength, but the overall quality is even better than I could have imagined so I decided it's honestly good enough to put as a 1.0 release as most changes to this are fine tuning at this stage. There is no longer a trigger word however you can now better positive and negative prompt things such as runes, types of weapon, placement, details such as anime, painterly, styles, backgrounds, depth, film grains, camera positions and more. The most major issue I found with this version is it's a little over tagged so I would suggest running it at lower strengths on most models (.1-.5-6) until I push a 1.1 with better data and tagging. Thanks for the support!

Files with SHA256: faa5d8790e75c7cce39fa5e53d45c669ad10fc10b21833f3e3f05baf0bc63297