I wanted to test the abilities of what NovelAI can do in terms of reproducing Photography techniques and styles. Anything from old techniques as well as modern and fad ones. In this experiment I tried out a total of 228 different styles with the exact same prompt and the exact same settings in terms of seed, scale, steps, options and output size. I’ve used a multitude of sources to gather the different techniques; my background in photography, wikipedia articles, “top ten” lists from all around the web and some friends into professional photography.
I generated all of these in one big batch over the course of a few hours. I chose the prompt based on what would yield a picture that was exact enough to be repeatable, but still not influence the effects too much. For instance I made no mention of hair color, as I noticed that if you color for specific subjects they override things like Black and white and Sepia.
All images have been made using the same settings and prompt:
Photograph of a beautiful woman, small breasts, intricate dress, soft smile, long hair, eyes, photorealism, {{{{______}}}}
Model used is NAI Diffusion Anime (Full), Steps 28, Scale 11, Quality Tags On, Seed 2239508625, Undesired Content Low Quality + Bad Anatomy.
Many of the techniques and effects yielded no particular effect, so in the end I ended up with 116 images out of the 228 prompts. The images that didn’t make the cut were either all almost exactly the same as the ground truth image or didn’t represent the technique at all.