Are you tired of Flux.Dev's plasticky skin? Well this is the checkpoint for you! Get the benefits of the Flux Krea finetune, without the censorship.

The checkpoint is marked as a "Merge", but it's actually a mix of both. I first trained the Flux Krea model on a dataset that only included nudes (no crazy concepts like BJs etc), but unfortunately I was about to run out of credit on vast so I stopped the training prematurely (about 10 epochs in). The result was an uncensored model that rendered nipples and vaginas properly about 50% of the time. So I decided to merge the PussyDiffusion and NippleDiffusion loras made by Skullkid, which ended up helping stabilize the model. (those two were merged with a strength of 0.5).

Text encoders and the vae are already included in the checkpoint!

Recommended settings:

Steps:: 20-30

CFG scale: 1

Sampler: dpmpp_2m

Scheduler: sgm_unifrom

All the preview images have the workflow attached, it's basically the standard Flux workflow with teacache added for extra speed, if you don't want to use teacache simply remove the node.