I've included a "wdgoodprompt" and "wdbadprompt" embedding in the embeddings folder to help make generation easier. With in progress models, its common to have to use long prompts for good results. Using these embeddings helps alleviate some of that.
With Waifu Diffusion 1.5, best results are generated from generating at a resolution of somewhere between 500 and 1000 and then using 2x latent upscale hiresfix
WD 1.5 is released under the Fair AI Public License 1.0-SD (
https://freedevproject.org/faipl-1.0-sd/). If any derivative of this model is made, please share your changes accordingly. Special thanks to ronsor/undeleted (
https://undeleted.ronsor.com/) for help with the license.