This is a merge for so many models that I have lost count of all the things in it. This creates beautiful semi-realistic backgrounds with ease. I highly recommend using <a rel="ugc" href="https://github.com/Bing-su/adetailer">Adetailer</a> and <a rel="ugc" href="https://github.com/mcmonkeyprojects/sd-dynamic-thresholding">Dynamic Threshold Scale</a> plugins to get the most of this model. I used Adetailer with default settings. Select any face model and enable (no keywords) and let it do its magic. Dynamic Threshold I mimic scale of 20. Simple prompts work best. I used some negaative textual embeddings in some of the example images but you dont really need those to get good output. Adding simple descriptions to background makes it detailed. Can be quite robust at generating NSFW content as well. Drop me a note/comment here what you think of this model with images of your generations.

Melancholic

Retains the beautiful semi-realistic backgrounds but makes the faces lean a bit more towards the anime side. Its more of a preference thing. As before <a rel="ugc" href="https://github.com/Bing-su/adetailer">Adetailer</a> is highly recommended to fix faces. I would go as far as to say that it is mandatory. It does take in some artist names as input. VAE is packet into the model.

Files with SHA256: aa1ef6cd1c6a780cc7e9ecb2397eae5aae39e5680894a364e583bbd78a2a0fa1