The model I think is fairly self explanatory.So I looked at the recent models uploaded over the last month for illustrious for checkpoints, and it didn't seem like there's many people doing semi-realistic / realistic checkpoints much these days, at least sorting by newest to oldest I didn't see too many. I don't think I'd be able to pull off doing a realistic model because that's several cans of worms, do you just build on someone elses work and work it from that angle, the alternative to synthetic data is real stuff, and that's a bit outside my wheelhouse. But I noticed that I had a lot of semi-realistic models of LoRA's, and it took me the entire day to develop a little bit of a clue of how merging LoRA's into checkpoints work now, though it generally causes issues where you do get a unique 'look', but with the interpolation you'll have to gamble on what data you're going to lose as you merge here and there, because you flat out wont know until you try prompting and discover that characters are incapable of putting their hands on the other persons thighs when they're 'on their knees'. Because that data isn't functional anymore. So this checkpoint likely has issues in all sorts of places since I generally don't know what I'm doing with throwing assorted LoRA at the wall and seeing what sticks and what falls out. Don't take this Checkpoint as a serious attempt to fill a niche, but maybe down the line it'll find itself in a merge someone else will make that I'll use for 'personal reasons' cough gooning. I do have some ideas though of making custom merges of some of the more notable artists who've done stuff that sticks out in my mind into a checkpoint and seeing what happens then if you merge this with that and then fraud. Just thought I'd try something new while the site is functionally broken for the day.Largely my previews were made with DDIM with DDIM, and then with hires fix DPM++2M and Karras, I used 40 and 20 hires steps for the previews, but honestly kind of anything beyond like 20-25/10-15 CFG around 2.3, but I think I was using a lot of addons that change noise and stuff, because I'm goofed up on coffee and I forgot I did that. Oh well. Basically everyone knows how to prompt better than I do, so I'm sure you don't have to worry about it. Just throw some spaghetti at the wall and see what sticks.Back to my devilish tricks.<a target="_blank" rel="ugc" href="https://vocaroo.com/1dCKgpymuk1M">https://vocaroo.com/1dCKgpymuk1M</a>[Malebolgia] Throwing Spagooter at the wall Semi-Realistic Checkpoint Illustrious

[Malebolgia] Throwing Spagooter at the wall Semi-Realistic Checkpoint Illustrious

MalebolgiaThrowingSpagooter_v10.safetensors

v1.0

The model I think is fairly self explanatory.The intention of this checkpoint is to be used in concert with Style LoRA's, while it's fine on it's own, it's true merit is in being a decent neutral checkpoint for Style LoRA's to work well with less effort and frustration by the user. I have provided 10 sets of 2 images, 10 have a 1:1 prompt including the LoRA of a model I've done in the past's cover image accompanied by a second image that is just THIS checkpoint without any LoRA attached. I believe all images were generated with DPM++2 at around 32 steps with 32 steps denoise-highres fix at a strength of 50 I think. I figure that's a pretty decent standard as a sort of legacy sampler you can easily use to compare between models and such. It's not personally what I use, but it works.small bloggingSo as you can probably see I was busy again for the last few days and vanished again like I do every month. I decided to embark on another one of these shots in the dark to create a checkpoint specifically for being used with LoRA's to generate images the way that the LoRA should look like in a neutral enviroment without the LoRA's weights being overpowered by the checkpoint Data, so I spent a bit of time trying to think of how I could make that work with A. Not knowing what anything does. B. Not knowing what I can use as 'neutral' art. and C. Somehow having success without being able to explain why it works the way it does that I was successful having learned that I'm pretty good at just randomly guessing how A.I works to make stuff out of it.<a target="_blank" rel="ugc" href="https://vocaroo.com/1dCKgpymuk1M">https://vocaroo.com/1dCKgpymuk1M</a>Back to my devilish tricks.Works well at a strength of 1Previews made with [Malebolgia] The 8th Circle - 'Fraud' - Custom Illustrious Checkpoint

[Malebolgia] The 8th Circle - 'Fraud' - Custom Illustrious Checkpoint

Keep in mind you will not be able to recreate the exact preview images of this model without special settings. Refer to <a target="_blank" rel="ugc" href="https://civitai.com/articles/26487">https://civitai.com/articles/26487</a> for information on how to replicate the previews. Also the other previews will have to wait. The site issues did indeed get worse. Hopefully they fix the bugs soon.

Files with SHA256: f6af9f4f4946eb1e6da35f288afe4422a31dd12383f707f337bb3b40f55b6948