This dataset was obtained by scraping messages from Midjourney's public Discord server over a period of four weeks (June 20, 2002 - July 17, 2022), from 10 channels. The resulting 250k user-issued text prompts, together with the URLs of the images generated by Midjourney can serve as an interesting data source for research on prompt engineering. The authors have no affiliation with Midjourney.
2 comments
[ 3.8 ms ] story [ 11.9 ms ] thread