UPDATE: After some tests today for a few hours, I’ve come to the conclusion that this version isn’t as good as it seemed at first. It has certain drawbacks that outweigh the benefits. I “poisoned” the model with one unstable merge iteration, and now it’s noticeable.
I’ll leave this model up as an archive version. It's still kinda workable and I might use it from time to time, but overall, I’m planning to create another, more stable version. Sorry for that T_T
You don’t have to download this one. But it's up to you.
DISCLAIMER: This LoRA does NOT improve working with ΣΙΗ - the base model is already good on its own. My LoRA can produce either a “better” result or awful one. Use it at your own risk! If you dislike unstable experiments, feel free to skip this - you’re not missing out on anything c:
This is a merge for ΣΙΗ. I built it specifically for this model, so it’s unlikely to work elsewhere. I originally made it for myself but uploaded it here in case anyone wants to try replicating my images.
This is NOT a style, but more of a direction for a style. I love the base model but wanted to add a touch of texture, strokes, and a bit of my own vibe. That’s why I created this schizo-merge based on my own LoRAs. And yeah, you definitely need to include artist tags. At least one. Please.
I threw this together pretty quickly, in about ten to fifteen minutes, so I haven’t tested it long-term yet. I might delete it, update it, or just leave it as is. We’ll see.
You can check some comparisons here. I might add a few more tomorrow.
Recommended weight is something between 0.5 and 1. Try 0.9 to start, then adjust up or down as needed for your workflow. At higher weights characters might lose some of their distinctive details. So be aware.
There are no trigger words required.
Good luck! You’ll definitely need it with this shit :^)
