An attempt to steer Noob1.0eps away from anime and more toward a Western cartoon style.
Recommended Reforge settings:
For images with detailed backgrounds use DPM++ 2M sampler with SGM Uniform scheduler, 24-30 steps and 3.5 cfg.
For images with solid backgrounds, use EulerAncestral CFG++ with Karras scheduler. 30 steps and 1 cfg. Image fidelity for sketchier styles also can be improved with APG turned on. Scale 0.8, Momentum 0, APG Threshold: 20, APG Eta: 1 (if your reforge version is newer than November 2024, these APG settings should be: Momentum: 0, Adaptive Momentum :0, Norm Threshold: 20, Eta: 1.)
Pony LoRAs work great with this merge and all Noob 1.0eps LoRAs will work as well.
Description
FAQ
Comments (6)
Hi, regarding this recommendation for APG:
"Scale 0.8,
Momentum 0,
APG Threshold: 20,
APG Eta: 1"
I have different settings, which are as follows:
Momentum (unsure of default, maybe 0.8?)
Adaptive Momentum (default 0.18)
Norm Threshold (unsure of default, maybe 16?)
Eta (default 1)
So there are two settings that have the same name as your recommendations, but they're in a different order, and I'm not sure if they actually control the same things; plus, two settings are named completely different. Do you have any advice for what the equivalent settings might be?
you're not the first to mention that your APG has different settings. I'm not sure which slider is equivalent unfortunately. I have a fairly recent update of reforge. I wonder at what point they changed those settings, and which ones are actually the newest version.
@someyoungguy69 As far as I know the settings I have are the most recent, looking at this commit from 24/Nov/2024: https://github.com/Panchovix/stable-diffusion-webui-reForge/commit/2ec985a48398c4639062c35612788284681fcc6b
"Fix APG adaptative momentum, also update APG to add guidance limiter."
I do indeed have both "Adaptative" Momentum [sic, typo] and the Guidance Limiter, which consists of its own section with an activation checkbox:
[checkbox] Enable Guidance Limiter (default Off/Disabled)
[slider] Guidance Sigma Start (default 5.42)
[slider] Guidance Sigma End (default 0.28)
Thank you for this model, worked great (with my limited experience with XL models).
best model for furry only content. I don't see myself trying anything else anytime soon.
I found this model surprisingly amazing to recognize all kind of concepts and styles, the backgrounds are amazing too. I would say this is one of the best I've tried