Abigail Brand, Director of S.W.O.R.D.
My first attempt at creating a LoRA for a character using a small dataset (16 low-quality images) and first Illustrious Model too.
Notes:
Use with 0.8 strength (higher may cause artifacts).
Use at least the following trigger word:
abgelv1.For a more comic-accurate look, add:
green hair,{green eyes|sunglasses},bodysuit,gloves.Positive prompt:
masterpiece, best quality, amazing qualityNegative prompt:
bad quality, worst quality, worst detail, sketch, censor
Recommended settings for text-to-image:
Sampler: Euler A (Karras)
Sampling steps: 20
CFG scale: 5–7
Then, send your result to image-to-image with:
Sampling steps: 30
CFG scale: 5–7
Denoising strength: up to 0.75 (lower values for a closer match to the original image)
Extras: ADetail (for face and eyes)
Created with: WAI-NSFW-illustrious-SDXL
But feel free to try other checkpoints as well.
I hope you will enjoy using Lora. Don't forget to rate it, leave comments, and share your creations! Your feedback is always appreciated! Thanks!