This model aims to reproduce the unique art style seen in the eyecatches from the 2007 anime "Tengen Toppa Gurren Lagann".
Important!
For the best results apply the LoRA on checkpoint AnyLora
Try strengths: between 0.5-0.9
If you're looking to reproduce some of my examples, don't forget to update the LoRA filename in the prompt. These were from training and I forgot about them.
For more details see https://github.com/jamesz96/ttgl-eyecatch-LoRA
Description
FAQ
Comments (3)
Results are not as good on other anime models and straight up damning on non-anime ones.
Anyone have any advice on this?
I tried training on the SD 1.5 checkpoint with the same relatively small training set (28 images) and couldn't get anything matching the aesthetics I was looking for.
Don't know what you're talking about. I used Based65 and it worked just fine.
From what little I know about LoRA training, you can do it with 28 images and still be fine if your tagging is good. Maybe to help reinforce the style you're after you should add images that have the look you're after, but not necessarily the pose? I'm not sure if that will negatively effect the "Eye Catch" part of this LoRA though.
Try using NAI (anime-full-prunned) as a base model instead of SD1.5. Most anime checkpoints are based on NAI. And probably most anime LoRAs you can find here use NAI or a variant for training.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.