This lora was trained with AI toolkit.

First I trained a realistic version of this Lora but it sucks because it is so hard to find high quality images in this pose, the good quality images you can find often have a bokeh effect. So the first attempt didn't get the pose correct 8/10 times. I took the good images from the generations and used them to make a new version which gets the pose correct most of the time.

This version will get the pose correct 8/10 times

Every image shown was made using the Deis Sampler and the SGM scheduler. Guidance 4, 24 steps. Use high strengths with the lora, .75 or higher. You can go lower and you will get less of the style shown but less consistent poses.

It does not do realistic images that well but I also didn't try that hard to make realistic images.

This version works pretty well but you have to at least write something coherent.

If you scroll through the display pictures you will see two raw batches of 12 images. No other lora was used for ether raw batch. These were the prompts:

1. High depth of field 24K UHD raw photo of a cute young Asian woman. She is lying on her stomach with her hands supporting her head and her perfect feet playfully raised in the air. She is wearing black and white sweatpants and a matching low-cut T-shirt.

2. A vivid raw photo of a semi-realistic flawless goth girl. She has girly facial features. She is lying on her stomach on her bed with her legs crossed behind her and her feet raised in the air. Her bedroom has cinematic lighting, with gothic decorations littering the walls. She is wearing gray sweatpants and a black shirt. Her arms have intricate tattoos of a gothic nature. She looks at the viewer with an annoyed expression. Her pale complexion and striking, dark, she has long voluminous hair styled in a layered, tousled fashion with black roots and white tips. Her eyes are large, almond-shaped, and accentuated with heavy black eyeliner and mascara, giving her an intense, mysterious gaze. She has a small, silver nose piercing on her left nostril and a small stud in her lower lip. Feminine yet intimidating features.

The example image were made while using this checkpoint: CG concept style by AIhuihuahua or the regular flux dev fp16

There is no trigger word. I only use Comfyui so I can only speak to this lora's performance in Comfy