A LoRA designed to help style amateur videos with unique options for body types, clothing, and small details. Trained off of an old dataset of from a FLUX LoRa I left unfinished and later used on a Hunyuan Video lora. I didn't change any of the captions from the FLUX dataset. This LoRA is INTENDED to give a slightly lower quality output more similar to cellphone photos in general. It also skews Z-Image away from it's default Asian tilt. Main Trigger Word: [amat_petite] (trained with the [] so they may be necessary) but honestly sometimes the output of the lora is better without the primary trigger word. See all training captions here: <a target="_blank" rel="ugc" href="https://files.catbox.moe/rwj3po.txt">https://files.catbox.moe/rwj3po.txt</a> PDF showing the same prompts without the lora, with the lora without a trigger word and lora with the trigger word: <a target="_blank" rel="ugc" href="https://files.catbox.moe/lzt2ue.pdf">https://files.catbox.moe/lzt2ue.pdf</a> What it's capable of:<ul><li>Amateur photo styles such as cellphone video and selfie angles</li><li>Petite body types<ul><li>Skinny, thin, short, tall, can get others out of it but it will skew towards petite.</li></ul></li><li>Wide range of breasts<ul><li>sizes but skews towards small breast sizes</li></ul></li><li>Various poses<ul><li>Kneeling, Standing with arms above ahead, reclined on bed/couch, running hands along body through hair, squeezing/holding breasts, hands on hips, general photography model posing, side profile</li></ul></li><li>Tan lines<ul><li>generally good at tan lines, lots of examples in the training set</li></ul></li><li>Braces<ul><li>The original dataset had a few training images with braces and honestly they work pretty well</li></ul></li><li>Clothing in pulled down/up states<ul><li>Not always perfect but works decently</li></ul></li><li>Different lighting types<ul><li>Best at natural light, should understand: "outdoor light, natural light, indoor light, overexposed</li></ul></li></ul>Sometimes causes weirdness with nipples but re-generating or changing the strength can help combat that.

Edge of Reality [Amateur Styles, Body Types] - Z Image Turbo

Trained at 2500 steps off of 154 videos. Trained to 3000 but found it to be overtrained so I picked 2500 as the best

Files with SHA256: 5e566c3d9a7a2815c612883d30b560045f1d446ffab747c997993b0f2cdfa807