This is my first style lora for Z-Image turbo so go easy with me on this one.To use, I recommend passing a reference image to <a target="_blank" rel="ugc" href="https://huggingface.co/spaces/fancyfeast/joy-caption-beta-one">https://huggingface.co/spaces/fancyfeast/joy-caption-beta-one</a> and generating a "very long" caption, as that captioning model was what I used to train the data. Prefix it with "boerbara" as the trigger word. I find the following works best: "boerbara. This is a highly detailed, realistic digital painting of a muscular anthropomorphic", followed by the joy-caption from the character species onwards. Upper body shots seem to work best also but it can do full body. I wondered what would happen if I threw all my bara training data and images into one Z-Image turbo lora and cranked up the steps and network. This is the result after many hours of training. It uses the new de-distillled training model recently released and is trained on over 350 images of muscular furry dudes. It's not perfect, extremely inconsistent and can't do peens. Honestly, it's more of a character lora than a style lora and I doubt anything will be that great until a proper furry finetune model is released one day. But hopefully people enjoy this in the meantime.

Files with SHA256: 5f144f38be95e509c16a3d6fb4a05b8d9c6da6fe3859660927e9dd0a8df47c0b