This is my First LoRA.

It is as experimental as I am inexperienced.

The LoRA was constructed using AI-Toolkit from 20 DSLR photographs (downsized from 8k to 2k resolution). The photographs were made by taking pictures of a painting canvas that was 'painted' with light caustics (LED Christmas tree lights shining onto the painting canvas through various pieces of glass and lead crystal...glass shadows, if you will). The images also included some glass pieces (crystal balls and a glass candlestick).

I wanted to make a LoRA that brought more abstract elements to generations. Brilliant and vibrant colors, with light caustics and textures enhancing the generated images.

I use ComfyUI and load the LoRA with strengths from .4 to 1. Results with strength values over 1 can be very interesting.

The LoRA can add abstract elements of color, textures, flares, and can add glass objects.

The LoRA is also good at adding abstract colored backgrounds.

Trigger words:

"GlassShadowsSpheres", "GlassShadowsSpheres style", "GlassShadowsSpheres wings", "Glass Candlestick", "Glass Spheres", "Crystal Ball", "Light Caustics"