PLEASE NOTE: The images you will receive when using this model unmixed will not in any way, shape, or form resemble the images provided here, which were generated by mixing this model with RealisticVision1.4. The images you will receive are, come to find out, against the ToS. (Though are not particularly gruesome; mostly they are just abstractly unsettling.)

I trained this on 640 photographs of crime scenes from the 1910s, taken from the NYC municipal archives. My intention was to make a thing that would generate random crime scenes for a noir/detective project without any need for prompting, and in that, it succeeds, though as a stand-alone, it is very much a one-trick-pony. It is very, very, very strong and inflexible—prompting is nearly useless—but I have mixed it with a few other ckpts to good effect (~30% of cr1m3 in the mix for best results, in my testing.) and you are encouraged to fuss around with it.

NOTE: If you see long crutch-looking metal poles around the edges of your synths, this is because many of the photos in the training data featured these. They're tripod legs.