Type | |
Stats | 909 1,008 452 |
Reviews | (214) |
Published | Jan 30, 2025 |
Base Model | |
Training | Steps: 6,650 Epochs: 100 |
Usage Tips | Clip Skip: 1 Strength: 0.8 |
Hash | AutoV2 D2C694DDFC |
Note 2-28-25: I re-trained the model to have less steps and epochs. As a result, it's a bit less overtrained and may yield better overall results. I'd recommend V2 over V1 based on some quick tests.
This LoRA is a bit experimental, but I still think it's neat! Initially, I was asked if I could do the underwear in the style of the High School DxD eyecatch screens. However, after making some changes, the model evolved into an official illustration style LoRA. This model was trained on the 80 eyecatch screens from the first three seasons (Heroes had a drastically different art style), as well as about 50 other high-quality illustrations made for the series.
Obligatory self-promotion: consider supporting or commissioning me at https://www.patreon.com/MetalChromeX or http://www.fiverr.com/s/VY34o4m so I can continue making cool crap for you all to enjoy for free!
Notes/tips for usage:
Keywords eyecatch and commercial break will help to more likely trigger the eyecatch/anime logo; the logo appears to randomly trigger, so you may need to try multiple times to get it.
Some characters are well known enough to Illustrious XL (like Rias and Akeno). I've also manually tagged characters; check the trigger word section for these.
If using a GUI, I recommend ADetail