Update for v1.4: Full model/weight changelog at bottom of Description.Join my new community on Discord: <a target="_blank" rel="ugc" href="https://discord.gg/xfCftj9pNE">https://discord.gg/xfCftj9pNE</a>Get in touch with me: <a target="_blank" rel="ugc" href="mailto:Mephist0phel3s242@gmail.com">Mephist0phel3s242@gmail.com</a>Or buy me a coffee: <a target="_blank" rel="ugc" href="https://ko-fi.com/mephistophel3s">https://ko-fi.com/mephistophel3s</a><h2 id="this-is-a-free-and-open-source-model.-3ibuveogx">This is a free and open source model.</h2>Workflows available as meta data embedded in all showcase and posted images.All generation meta available directly in pics.Update: Devil DiffusionSee my GitHub.<h3 id="https:github.commephist0phel3sdevil-diffusiontreedevil-qqok6shf2"><a target="_blank" rel="ugc" href="https://github.com/Mephist0phel3s/Devil-Diffusion">https://github.com/Mephist0phel3s/Devil-Diffusion/tree/devil</a></h3><h2 id="overview-of-devil-v1.4-yl86c0plm">Overview of Devil v1.4</h2>This Merge is a combo of many other models and ideas, most notably:Realistic and semi realistic CartoonRealisticSplash of anime for style and flavor.Optimized to run on Linux. Runs well on NixOS.Unknown if potato PC's can run this model, please give me some feedback if you got a potato, though i dont think its much of an issue.Usage Tip: This model has been rigorously tuned mostly for Euler, aligned steps scheduler type, and high step counts.Example.Euler - Align your steps OR Simple OR Karras - step count 60+ for best results, more steps will refine detail in the background and around tags for "perfect" or "beautiful" or "flawless"Works aight with other non Ancestral sampler types.Not really tested much on Ancestral scheduler types, please post feedback on other scheduler types if you are interested.<h2 id="-34jyo9n7v"></h2>No suggested prompts, model should be flexible enough to do pretty much whatever you like.<h2 id="this-is-a-free-and-open-source-model-expanding-on-the-idea-behind-the-devil-sd-1.5-model-i-previously-released-but-this-time-using-pony.-still-under-development.-dc844oheg">This is a free and open source model expanding on the idea behind the devil SD 1.5 model i previously released but this time using Pony. Still under development.</h2>This model will produce good excellent quality NSFW images, this model excels in realistic and semi realistic cartoon in particular, and does a fair job at artsy styles and Anime.The main focus of this model was to produce a blend of elements that would produce quality images of Devil girls and Succubi in particular. Such as Meru the Succubus.This will be an open source model, anyone can take and use this model, merge it, tinker, etcI only ask that any and all merges, and/or content is held to the spirit of open source.You may sell works derived from this model so long as you are not using another resource alongside it that disallows this.I also ask that any merges made from this model also follow this guideline and be made freely available to anyone either here on civit, or somewhere else publicly available without restriction.I want anyone to be able to use my models freely without fear of retaliation so long as the spirit of this request is upheld.Though being credited in any works/merges yall might produce would be greatly appreciated, it is not necessary.The only thing i ask in return is a star on some of my work and perhaps a tip. ;)<h3 id="v1.4-changelog:-1mme6n7w5">V1.4 Changelog:</h3> 40% merge towards ultra realistic on **all blocks**.Illustrious weights + SDXL + Original Pony weights. More IL now than not.

Devil Core

After several days tinkering with the weights, ive finally gotten the model to generate realistic textures but add just a touch of anime/drawn to the detailing and pose generation.It almost feels like a mix of ultra high res 3D + anime feel at the edges but still retains the core of the model which is to generate realistic images.Euler + simple, karras, or aligned step counts are recommended as this model has been rigorously tuned for this. Other types are mostly untested with this release as they are not a focus at this time.This model comes with a VAE baked in.This model does not require Clip Skip to generate good images.Anime and realistic loras will both mix well with this version.A workflow for Comfy ui and the exact generation methods are embedded in each img in the showcase. Drag and drop to load the workflow.

Files with SHA256: be416b90f60ded387c3a8e36015954e67f81a0c47f4112fd6d60e4fa4c72cd1d