Trained on novelai on 96 images (1024x1024) of Secre Swallowtail with network rank and dim: 32/16. More flexible and more accurate now than older version. It works without using the trigger word, but I recommend using it anyway.
Use red eyes if eye color gets wrong.
Description
-v.1.0
FAQ
Comments (8)
Love your LORA and Secre, was waiting for someone to make one of her. To be honest your Lora looks really good.
Hey, quick question, what settings did you use to train your LORA? I always train full CKPT files but i'm thinking of getting into the LORA space but i'm not quite sure which settings to use. What settings do you reccomend?
Thank you! This is my first LORA! :)
I used kohya SS Lora. Just follow this guide, it is close to how I did. https://www.youtube.com/watch?v=70H03cv57-o
I also used a 768x768 resolution. Going to try higher next time.
@Robeloto No problem, thanks for the info. Keep up the good work!
@Quiet_Joker Thanks :)
Hey, I like your Lora but ill give you a small hint for training
Dont use abyssorangemix or anything else use NovelAI checkpoint for training its better
if you want the file add me on discord: nicerpanda#8302
Hi! I will try out the NovelAI checkpoint! Thanks for the tip! Will the quality increase with NovelAI? Never used it before I think! ^^
@Robeloto With NovelAI the versatility of the LoRA will increase, so it will properly on more models than just AbyssOrangeMix based ones
@LUUNA Thanks for the info! I will update this model soon using the NovelAI checkpoint. I also managed to get a hold on higher quality images. So hopefully it will turn out ok.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.