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

Couldn't put off fixing the model, after using it in depth over the last two days I sort of puzzled out a few things that were wrong with it, such as needing to use LoRA's at a strength of 2 that completely defeats the logic and purpose of the model. That should be resolved now, and the model should function as I originally intended. I'll update this in the future if I come across issues that I notice again like I did with the prototype. Preview images for 'The First Ditch' of Fraud are remakes of the models I did images of with 'Fraud' to showcase that the damn thing works this time with less hassle.

Files with SHA256: 030511f02dd17f89c383efbec20a38148a51036047fc7e694f4e57866a733176