You may notice that there are two versions available. Which should you use? While v1 can respond better to more complex prompts, v2 works with simpler prompts and just adds detail and color to them. v2 had a larger dataset than v1 so it is more diverse as well. I would recommend v2 for simpler prompting, but you can always use both!
This embedding can be used to create images with a "digital art" or "digital painting" style. It does portraits and landscapes extremely well, animals should work too. At least the well known ones. Architecture is ok, especially fantasy cottages and such. Sci Fi is probably where it struggles most but it can do apocalyptic stuff. It can also sometimes do anime. Feel free to experiment with these and anything else and leave a review to share your findings!
Try this AI prompt generator (a standalone program, not extension) made by 526Christian and I. https://github.com/526christian/AI-Image-PromptGenerator. It is basically an advanced version of wildcards, and it works super well with this embedding!
Because this is a 2.1 embedding, you need to use a 2.1 model. (It might work on 2.0, I am not sure). I have only tested it with the base SD 2.1 512 & 768 models but any other model should work as well, but it could not be like the example images depending on what you use.
Please share what you make, I would love to see it!