*******VERSION 2 UPDATE*******
Version 2 is an upgrade in every way, as version 1 was considerably overcooked. It has two styles baked in; "anime screencap" for a more powerful Amagami SS anime style, and "promotional artstyle" for the glossier art they used for the promotional material and extras. If you want the "promotional artstyle" look without the gloss, add "anime coloring".
**************
First attempt at an Illustrious LoRA since everyone is playing around with it now. I will be the first to admit I have no idea what I'm doing. I used the same dataset from my pony version, but I have no idea what kind of parameters I should be using for Illustrious training.
To be honest, I don't even know if my prompting for Illustrious is solid. While it seems much more responsive to prompting than Pony, I keep getting these crazy artifacts and I can't tell if it's a result of the model or from my dataset.
I figured I would upload the attempt, might re-attempt or attempt other Illustrious LoRA once there's some more info. I'm pretty much making a stab in the dark here
Description
FAQ
Comments (2)
Your LoRA works great! I do not see the problems you mention.
It works best at a low weight such as 0.4~0.5.
This might be because the LoRA has too many steps. You can try earlier epochs at higher weights (0.8~1) and see if you get the results you want.
Works great. Looking forward to future Illustrious loras
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.