Been tinkering with this for a while. Still trying to figure out what prompts work best, since it's a merge of a lot of models.There's also the anime-style version of this model (more like 2.5D) <a target="_blank" rel="ugc" href="https://civarchive.com/models/30382/mcq-ani">mCQAni</a>I'll update the description if this goes anywhere, including the recipe.I used <a target="_blank" rel="ugc" href="https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt">vae-ft-mse-840000-ema-pruned.ckpt</a> for all image gens.Pruned Model fp16 has baked in VAE.Might be a good idea to include sfw/nsfw respectively into prompt/negative prompt depending on what you're looking for.(note, I'm running macOS 13.3, python: 3.10.10, torch: 2.1.0.dev20230326 as of posting this)Update (Apr 14):Seems to work best if you include: "((masterpiece, best quality)), highly detailed, photorealistic, photo of" at the beginning of the prompt, good results with "((perfect face, detailed iris, realistic pupils)),(high detailed skin:0.9), (visible pores:0.3)," at the end.However, you should get good results with simple prompts as well. Check example images, there's a variety of samplers, prompting styles, with and without hires fix.

First Public ReleaseMix contents in short:<ul><li>schoolAnime</li><li>AOM3A3</li><li>AnyThingv4.5</li><li>koreanstyle25D</li><li>camelliamix25D</li><li>3moonREALCartoon</li><li>level4v5</li></ul>I THINK (kinda new to this) there's no vae baked into the big model.

Files with SHA256: bf9b9678bbe2873e84ce44a7e5a73fd5244eda324ce89e0e463b20ab685869c5