The first of a series of art-style LoRA's for Flux.1-Dev. The approach is based on mnemic's experiments with not using a caption at all. It seems to work very well for simple art styles, where the effect is applied to the entire image, rather than just a portion.

Further experimentation with single-image training underway.