Norman Rockwell style LoRA for Pony V6 XL and Illustrious-XL v0.1.
For best results, prompt using Booru tags.
See training data for image list, tags and training config.
Description
Recommended sizes: 768x1024, 1024x1344.
Used 41 tagged/labeled images, mostly bucketed in 768x1024. I didn't attempt to remove "The Saturday Evening Post" text or signature.
As a test, I only trained on images I considered "minimal colored background", and excluding "detailed background" and "minimal neutral background" as I collected a few hundred of these in total.
FAQ
Comments (3)
Not QUITE Normal Rockwell. But what it is, is quite nice.
I feel like there is too much of an asian influence in your output.
Yeah I agree, V1 is a bit weak. I'm investigating using larger LoRAs and labeling his "detailed background" works for training.
I used the same script as "V6" to train "V12" on Animagine XL V3.1 which (I think) now matches his style (realistic) more closely. However, I assume because Animagine is more similar to the base SDXL than PonyV6, it is also easier to get bad results, at least with this LoRA alone.