This repo contains 100+ prompt only photographic style references for SDXL models, optimized for RobMix Zenith. I created these as a study in how my model represented certain concepts. I haven't tested these with other checkpoints.
Styles are fully "public domain" without reference to specific artists, and were generated from reference images using ChatGPT vision to extract the stylistic elements of each, then fine-tuned for the desired result.
Click the image thumbnails to embiggen. My favorites are at the top of the list marked with an asterisk.
These styles will work by simply adding them to a prompt, but many are too long for CLIP's 77-token limit. I recommend using conditioning concatenation. See example workflow for details.
Some styles may be a bit heavy handed, and you may need to adjust weights in your subject or style to produce the desired image. Use these as a starting point for experimentation.