DreamVerse

Hello and welcome, thanks for checking out my very first attempt of a checkpoint merge. It might not be the most innovative or original one out there, but that wasn't the goal here. And for what is essentially an experiment it turned out quite okay I'd say. Keep in mind, this is version1. If people want me to (and if I figure out how to), I will try to improve it of course. Creating pictures and sharing your prompts would obviously help a lot, so feel free to leave a review and give me your honest opinions.

Credit goes to Lykon

and wrench1815

for creating awesome models we all enjoy. DreamVerse aims to bring you gorgeous people (in gorgeous clothes), while also aiming to create decent landscapes etc., and all of that in different styles.

After a couple sessions of testing, I can at least say this:

• DreamVerse works well with BadDream, although it doesn't always seem to be necessary. Especially when you try to experiment with different styles, I have found that excluding BadDreams yields more accurate results. There is still room to experiment.

• Adetailer can be of use, but it is also not always necessary (at least for faces/portraits and bodies). Hands appear to be a bit of a weakness (but when are they not?), so I still recommend using Adetailer + appropriate embeddings there. The reason why I don't always use Adetailer for faces (especially for portraits) is that it tends to make all faces look almost identical. This problem can probably be circumvented by using an Age Slider Lora or different prompts. When you create full body pictures, Adetailer is recommended.

• As mentioned above, it works wonderfully with most styles (tested with StyleSelector) and you should experiment with different embeddings like BadDream to see how, or if, the results differ.

• Responds well to prompts like "vibrant colors". CLIPS Skip 2 is good, negative prompts like "watermark, text, signature" are recommended.

With all that being said, I'm sure I forgot to mention something!

Either way, I hope you can enjoy this model, and maybe together we can even improve it. Suggestions, thrown tomatoes, all is welcome. Again, this is an experiment.