unstableJugg | Artistic LoRA Lover | Merge of Juggs and UnstableDiffusers版本unstableJugg v1.0 (ID: 658613) 综合资源合集综合资源合集

Hello everyone! First checkpoint to share with you all. It's a merge of some of my favorites! Original credit goes to Yamer and KandooAI for their work on the original models. Be sure to check out their work if you haven't!

Please follow all guidelines for use that were stated on the original checkpoints' pages.

Good settings: CFG 4-7 | Steps 18-22 | euler normal or dpmpp_sde_gpu karras

Use a refiner if you want to, or don't. I use one out of habit but don't use one when working in Krita.

What's in it?

Juggernaut v3 & v5

Unstable Diffusers v5 Unchained Slayer, v6 Stability Eater, v7 Electric Mind, v8 Heavenswrath

Uses VAE and Clip from Unstable Diffusers v8 Heavenswrath

What's a good prompt?

I'm a bit of a minimalist prompter. I don't really use negatives unless I need to. I usually do live painting in Krita to fix stuff. Short prompts seem coherent to me. Longer prompts, or the ones that are entire chapters from a book, it's hard to tell if you really get what you want. However, seems like they still turn out nice images. A decent amount of images on this page are prompts taken from the original models. The ones that are super short, those are mine, aha.

What can it do?

I can't really tell you that. I recommend checking out the original checkpoints from Yamer and KandooAI. Look through their work and that will give you an idea of what "could" come out. Remember, this is a merge, so some of their original work is washed out in certain parts. Use their original checkpoints if that is what you are looking for.

Why though?

As you might be thinking, why would you want to download yet another checkpoint, that already has awesome checkpoints in it that do fantastic on their own? Well, I love each one of these checkpoints a ton and use them a lot. I found that I would be switching between all six of these checkpoints to make what I wanted and if you know how long it takes to switch checkpoints, every time, that a lot of time spent loading. There are pieces in each checkpoint that I love and parts that just work fantastic with certain LoRAs. I mainly use these checkpoints with LoRAs and when building certain piece of AI art, I would spend just too much time switching. The developers would come out with new versions and I would think, this one... this is the one that's going to please me with all that I want. After downloading and downloading, I kept coming back to these six checkpoints for certain and many applications. So I finally researched how to merge checkpoints together and started experimenting.

These checkpoints were merged together using the simple merge block method. The node just uses ratios. The ratios were 50:50 across the board, washing out each checkpoint one by the one piled on top. unstableJugg has benefits and losses from the merging. There are certain things it isn't good at that you are better off using the original checkpoint from Yamer or KandooAI. But for my purposes, it meets and exceeds what I wanted. I now can use the one checkpoint to do what I want and not shuffle loading different humongous files.

What is the strength of unstableJugg?

unstableJugg handles world morphing LoRAs very well, has good quality, and compositions I enjoy. I don't know that unstableJugg makes very many images that I don't like or can't fix up through live painting in Krita. It has that artistic wildness that can show up, depending on your prompt, like how Yamer put into their Unstable Diffusers checkpoints. I believe that the Juggernaut v3 & v5 bring the LoRA handling that I liked so much from them. The character LoRAs that I make tend to be tricky to stay consistent with certain checkpoints, so I have aimed at merging certain ones together to get models that perform how I want them to.

TLDR; It makes stuff. I like the stuff it makes. It's one of my flavors. It's a creative artistic style, though sometimes pleasantly simple.