I was looking to create an easy out of the box way of either generating new images, or edit existing images, easily to do the "Guilty Challenge" pose that was all the rave a few years back (don't ask). I had three approaches - 1) do it natively with Flux 2 Klein with use of prompting and ControlNets, 2) create them via another model that has existing LoRas that are currently available and then have Klein transform them into realistic higher-quality photos, or 3) create a Klein LoRa that does it. I went with option 3 since I thought the other two approaches would eventually be as much work as the third.Well, I was wrong - since I only wanted to do this for a handful of images it turned out creating the LoRA was more difficult than expected - it suspect because I'm fighting with Klein's inherit knowledge of clothes, arm poses and SFW male anatomy. It will tend to generate the forearm over the shirt though almost all despite prompting specificity - or do it with the forearm under the shirt but hand not visibly touching the face. So the LoRA has to fight against those tenancies which also requires finer concepts like not generating a shirt sleeve and keeping the shirt collar intact over the arm.This LoRa can generally do that - although it is not what I'd consider production ready stable. It'll do it maybe 25% of the time - so I ended up just having my batch count a bit higher and then chose the winners out of the batch outputs. Do not expect this to be a one-shot and done - it's highly likely you'll need to have it generate a handful (or two) of images and pick the winner or do a post i2i correction.I did throw in some NSFW concepts like peen and beans, bulges, some testicle peekage, etc, so it can be used with a SFW checkpoint and still generate NSFW content if so desired. Using it with a NSFW checkpoint like SNOFS does introduce quite a bit more stability, especially for the naughty bits, but I'm posting generic Klein 2 9B generated images in the image gallery to give a baseline understanding on what it can generate.Enjoy.

Files with SHA256: 9eab6612178c4bd902d844914e8d03306d4efd2e8eda937871eaf64b4815aa47