<h2 id="klein-9b-t2ii2i-erect-penis-model">Klein 9B t2i/i2i Erect Penis Model</h2>Proof of concept model for Flux 2 Klein 9B, trained on 9B-base with 30 images for 5k steps.Additional/more diverse images would help the model produce more varied results (it tends to prefer highly muscular physiques currently)Being that the base model can edit, this can be a step in your diffusion process as it will not affect other parts of the image if you tell it not to. If feet/hands are not present in the image in edit mode, you will most likely experience the flux anatomy scrambler and will have to edit the output image to improve hands/feet.The trigger word used for training is j0rkinit, it is not usually needed but can help in some cases. No explicit words were used for training as explicit words tend to create bad results. The strength is fine at 1.0, up to 1.2 can help with size. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/40a5258b-f3a2-46d7-99d7-42c9b50d358d/original=true/40a5258b-f3a2-46d7-99d7-42c9b50d358d.jpeg" /> The lora greatly benefits by drawing the general size/shape you want in the mask editor before generating, it gives the model something to structure the output around. The paint color doesn't matter most of the time, the model will occasionally use the color for skin tone. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/12dda501-73d7-4356-ae24-c1b35dbf0169/original=true/12dda501-73d7-4356-ae24-c1b35dbf0169.jpeg" /> It also works with illustrated images, you may need to specify "keep the art style the same" but for the most part it follows the content pretty well. DISCLAIMERBe respectful with edited images, do not share/generate images of real people.

Erect Penis Model (Klein 9B)

Klein 9B t2i/i2i Erect Penis ModelProof of concept model for Flux 2 Klein 9B, trained on 9B-base with 30 images for 5k steps.Additional/more diverse images would help the model produce more varied results (it tends to prefer highly muscular physiques currently)Being that the base model can edit, this can be a step in your diffusion process as it will not affect other parts of the image if you tell it not to. If feet/hands are not present in the image in edit mode, you will most likely experience the flux anatomy scrambler and will have to edit the output image to improve hands/feet.The trigger word used for training is j0rkinit, it is not usually needed but can help in some cases. No explicit words were used for training as explicit words tend to create bad results. The strength is fine at 1.0, up to 1.2 can help with size.DISCLAIMERBe respectful with edited images, do not share/generate images of real people.

Files with SHA256: 9c0a72831e4af4f1e1fbe47026471bd84082837ca68e4e5a269386df3028951e