It is a visual style LoRA.

The images in showcase are straight from training, created with an extremely simple prompt -

"A medium shot of a blonde woman wearing a streetwear hoodie, standing in a subway station, looking at the camera".

The ambiance and the vibe is entirely due to the LoRA.

Trained on TA on a dataset of 100 images, Captioned with Qwen VL;

Credit: This model received technical support from ModelScope, i.e. it was trained from scratch (including captioning using Qwen VL) on ModelScope platform.