With the spread of the use of Text2Img diffusion models such as DALL-E 2, Imagen, Mid Journey and Stable Diffusion, one challenge that artists face is selecting the right prompts to achieve the desired artistic output. We present techniques for measuring the effect that specific words and phrases in prompts have, and (in the Appendix) present guidance on the selection of prompts to produce desired effects.
@article{arxiv.2211.15462,
title = {Investigating Prompt Engineering in Diffusion Models},
author = {Sam Witteveen and Martin Andrews},
journal= {arXiv preprint arXiv:2211.15462},
year = {2022}
}
Comments
Paper submitted for Creativity and Design workshop at NeurIPS 2022. (4 pages including references + 7 page appendix). We would like to thank Google and the ML Developer Programs Team for their assistance and compute credits used in the experiments for this paper