Type | |
Stats | 1,140 |
Reviews | (229) |
Published | Aug 7, 2023 |
Base Model | |
Training | Steps: 1,500 Epochs: 6 |
Hash | AutoV2 02A46BB87A |
This is an SDXL LoRA trained on Tim Jacobus' gallery of Goosebumps cover art.
I use the LoRA at 1.0 model and clip strength for all gens in the gallery.
Try this prompt:
tim jacobus style painting of {a-tree-monster|a-pack-of-warewolves-wearing-gucci, prada, and fendi|many-skeletons|a-slime-monster|dangerous-beings|a-ghost|a-sexy-vampire} {doing-something-terrifying|killing-someone|eating-a-baby|starting-a-fire|causing-an-accident|doing-science-experiments|running-a-bank}:1.4, extensive color palette, balance of warm and cool tones, visually appealing, detailed shading and lighting, depth, soft shadows, bright highlights, Utilize classic oil painting techniques to paint a horrifying picture.
BREAK
volumetric lighting, diffuse lighting, ambient occlusion
BREAK
located {in-a-graveyard|in-a-basement|under-the-stairs|in-a-school|in-a-science-lab}
BREAK
film grain:1.4,
negative:
syndrome:1.4, naked:1.4, nipple:1.4, exposed-nipple:1.4, child, loli, kids, close-up, close-shot, cut-off, monochrome, watermark, bad art, poorly drawn hands, bad hands, out of frame,
Generation settings:
DDIM, DDIM_Uniform, 40 steps, no refiner, upscale 1.5x with Iterative upscaler using dpm++sde_gpu, 20 steps, with facedetailer (not really necessary as most of these output are horrifying anyway.)