Originally posted to HuggingFace by MoistMix
Beautiful lighting, paintings, portraits, multiple photography styles, photorealism, anime and animated styles, alien creatures, armor, clothing, massive dreamy landscapes, abstract retro art, horror, space, nsfw, extremely detailed close ups, and whatever your imagination brings (or AI generated prompts lol).
Combining about a dozen other models/mixes and a few trained on some of my own datasets, I hope you all love this as much as I do, it's the only model I've used for weeks now, slowly perfected into what it is today.
This model includes a CKPT and a SafeTensors filetype which you can use the same way as a CKPT, with (sometimes) increased load speed while avoiding pickles. I am including my VAE as well.
Description
Based on SD1.5
A much improved version of my last, quite a different style, but sticking to the same theme as a "do anything" type of model. It LOVES artist prompts and they can change up the whole image dramatically.
For any questions or feedback: Pixiv
This model includes a CKPT and a SafeTensors filetype which you can use the same way as a CKPT, with (sometimes) increased load speed while avoiding pickles. I am including my VAE as well.
FAQ
Comments (12)
why did you have to call it "moist"? Worst word ever
or best?
Because he is Charlie in disguise.
Lily likes that answer
Please can you give prompt for that monster with the mask. That's unreal, love it.
thanks! was kinda a happy accident :)
demonic, evil, nsfw, bound, tied up, restrained, (shibari), kinbaku, petite 1girl,science fiction, (beautiful eyes), prisoner in a dark industrial facility, photo realistic, 20 megapixel, nikon d850, ((full body intricate, vibrant, photo realistic, realistic, dramatic, sharp focus, 8k)), (faded freckles:0.6), subsurface scattering, sharp, retouched, intricate detail, by Greg Rutkowski, by Jeremy Lipking,:0.8), ((junji ito)), by ralph bakshi,((Silent Hill)), H.R. Giger, Beksinski.
Negative prompt: asian, cartoon, 3d, (disfigured), (bad art), (deformed), bad hand, extra fingers, (poorly drawn), (extra limbs), strange colours, blurry, boring, sketch, lackluster, big breast, large breast, huge breasts, self-portrait, signature, letters, watermark, desaturated, monochrome.
Steps: 15, Sampler: DDIM, CFG scale: 12, Seed: 4248511735, Size: 384x640, Model hash: 55ae9e60, Batch size: 3, Batch pos: 1, Denoising strength: 0.66, ENSD: 31337, Hires upscale: 2, Hires steps: 15, Hires upscaler: 4x-UltraMix_Balanced
@Moist Thanks so much for the prompt. Only just saw you replied now.
The VAE for V1 and V2 is the same ?
Can you show at least one prompt of a portrait for each version ?
I am wondering the same thing seems the uploader doesnt care to reply to and about their creation. smh.
I wish you can share prompts of your images. It will help us create new prompts to generate some fascinating and stunning images you created.
Any tips for using this? I can't get even remotely good images with it. Should we be using Lora's / Hypernetworks ?? Cause without them and using the standard VAE I just get awful images from this so far... which is a shame as I love the style of it.
V2 is more like a dough now, and each of us is the pastry chef, and this dough is hard, or at least harder than V1. To shape the bread, then the pastry chef needs to be solid enough, and it can, relatively speaking, be more rigorous in only thinking about the pastry chef's instructions, provided you have the corresponding mould. This is the reason for the polarisation in the comments section, where a baker with good tuning skills would prefer an obedient dough, whereas the V2 seems to be an unassertive, but malleable dough.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.