Initial training was for busty, large, neutral ???????, but also helps with thick people too...

Trained on carefully generated, edited and tagged dataset completely anonymized and by me.

My main focus train on such a way that it only affects the related parts of the image, no trigger word needed, just describe the ??????? and it will do the trick.

Here is example prompt to create similar images:

busty ???????. A woman with blonde hair and blue eyes, wearing a bright pink ?????? top with thin straps, is positioned in front of a white bathtub. Her head is tilted slightly, and she gazes directly at the camera. Her lips are coated in a reddish-brown lipstick. Her shoulders, collarbone, and upper ????? are visible, with the ?????? top's band resting low on her torso. Her left hand is partially visible near her hip. 

I used official comfy workflow. Didn't test them with lightning loras, usual settings are:

0.8-1.0 strength
Euler/beta
CFG 3.5 - 4.0
30+ steps