A general model with a slant towards classical paintings. It has some anime capacity, but it isn't a core competency.Includes lineage from MangledMerge, Cheyenne, FabledIllusion and c0rruptf33d, and technically manmaruMix. I used Hako Mikan's Elemental supermerger and tested large numbers of possible ways of combining the three primary models, using python scripts to attempt to pseudo train a good setup. I also pseudo trained in some c0rruptf33d, when I noticed the model was more versatile than it appears. Once I was somewhat satisfied, I decided to use all my buzz to train enhancement Loras to "bring together" the merge. The idea being that, the merge is close to correct on a lot of concepts, so with a little training, it should automatically select the proper model that understands the given concept. This, seems to have worked? At least if the concept is in the training data and one of the models, it seems to have picked up on that? At the least, the model seems competent for what it is, and I don't see an obvious way to improve it further, besides throwing more resources at it, at the moment.

Arc Naked Singularity

In general, the model seems more detailed and vibrant now. Used Hako Mikan Supermerger's Plus Random to pseudo train the model by generating lots of models and images, selecting the model seeds I preferred and updating slightly in that direction. Then did some more merging with Chattiori Feature Weighted Merge and Hako Mikan Supermerger Cosine Merge to try to isolate positive changes from random changes.

Files with SHA256: 037f63f60c635835ea3f5081b5c01234fff1da8a26417df8296148e8b7be3197