Generates images of women with micro bikinis. I recommend upscaling after generating, not all images used to train were high quality so might be a tad blurry. I used SeedVR2 to get the images posted.
Description
FAQ
Comments (8)
bro how do you generate so perfect prompts for these baddies?? and do you use org zit or any other zit model?? for me i use org zit Q8 gguf.....
I used to use qwen 3vl and would create an ultra detailed description of some of the images I use for training. For this lora i used joycaption since its more NSFW oriented and qwen3vl doesn't really like NSFW. I just use the original z-image and 9 steps with 1 cfg.
huihui_ai/qwen3-vl-abliterated:8b-instruct 这个版本的vl模型可以nsfw
@TheAINovice btw can you share custom prompts text for joy caption which you use (i tried joycaption using runninghub as there is no nsfw restriction regarding texts) also i used LIama 3.1 8B Lexi Uncencored v2 model and these are my settings
Caption type - Descriptive
Caption length - long
top_p 1.00
temperature 0.6
@TaiLong I'm not sure how to use custom prompts for joy caption. I run it using comfyui and do
model - joycaption beta one
quantization - balanced
prompt style - Descriptive
caption length- very long
I'm not familiar with Lexi Uncensored v2. Is that used run locally?
@angle2043 didn't know this was a thing. I'll have to try it out thanks!
@TheAINovice i dont know , i just used it on running hub and i used joycaption2, btw can you share yours joycaption workflow
@TheAINovice 这个插件可以调用ollama或其他平台的API来生成正向提示词，也是有英文提示词输出的：https://github.com/yawiii/ComfyUI-Prompt-Assistant