DoRA Weighting Note: SDXL Illustrious DoRA weighting begins at 0.0, with 0.05 - 0.4 as optimal range. Start at 0.0 and increasse by 0.01 to desired level.
This is a style model based on artist as determined by BLIP captioning.
This is a separation of BLIP caption style from the "Style Catalogue". The Style Catalogue will focus on CLIP based captioning. Artist of SDXL will focus on Deepbooru/BLIP captioning. BLIP labels images based on artist, focus, art style and school. Variations , but basically the style is broken into 4 categories.
These models will focus on artist and art style. BLIP associates all images with an artist, allowing a more focused style model.
Description
create realistic fantasy painting scenes and characters
BLIP Artist Style Type: Amanda + (Sage) (loosely) The dataset does not have prominent psychedelics' visionary art similar to "Amanda Sage" Suspect weighting of "Sage" in the name drives the dataset. similar to "Alex Grey" leans to nsfw instead of psychedelic art.
Art Style: fantasy, painting