V1.32This version continues to improve quality, and give more consistent results. In testing with a lora strength of .6 it has seemed pretty easy to get images with a good stillsuit. This version has no masks. It will also do children pretty well. Trained resolutions 832x1216 mostly.V1.31The stillsuit's are much higher quality in this version, and take color much easier. Has better male resources. withoutmask prompt has been removed (it is in this version, but doesn't work. I plan to remove it in future versions).Suggested positive prompt: human head, natural skinSuggested negative prompt: long torso, long neck,v1.3This version finally brings withmask and withoutmask prompts. This model still needs a lot more images to train on, and is a WIP.v1.21It should have a sharper stillsuit, and the abdomen should be more consistent. It was also trained without cloaks. Using stillsuit with tattered cloak works pretty well if you want to add a cloak. Mask up and mask down will, hopefully, be worked out in future versions. For now, it is just luck of the draw.v1.1This is trained with 2 triggers, stillsuit mask up, and stillsuit mask down. Use either prompt, and expect erratic results because you could get either version, but more than likely you'll get mask up. The quality was a big enough improvement, so I am publishing it while I figure out how to improve the reliability of mask up/down.v1.0Lora of a Dune Stillsuit. The second version of this, v1.0, has quite a bit better quality than the first, but still needs some improvement. Keyword is stillsuit. Be gentle, this is my first time training.
