The name comes from the intentions I had creating this model: I was basically just throwing checkpoints I had in my folder together to see what came out.

From my tests with this merge it seems to play well with LoRAs. Plays nice with NSFW as well.

I recommend playing with the sampling method to get optimal results, I've gotten good results with both Euler variants alongside the DPM++ samplers, both Karras and non-Karras variants.

Recommended step count is somewhere from 40 to maybe 60, though I also suggest experimenting with the steps to see what works for you and your chosen sampler. The pictures posted used Clipskip=2 but it seems like the merge looks good with Clipskip=1 as well. It's all up to your preference which one to use.

The VAE is baked into the merge. No external VAE required. (Ignore the VAE recommended pop-up on the side, I don't know how to disable that. I apologize.)