V2

Updated dataset and trained locally. Now it should produce cleaner and better results. Base image is still changed a lot at weight of 1 so feel free to play with it. All principles of V1 apply here.

Dataset was expanded, so now it has more colors baked in and not that tilted towards neon colors without prompt.

Also due to expanded dataset and better quality I qualify this lora as a general glow enchancer

I was not able to get FLUX to draw abstract spirals of light the way I wanted so I made a LORA for that. It is good at abstract vortexes, where FLUX is always trying to create some sort of scene, try it out.

Have fun!

Activation tag: spirals of light.

Lower the weight to get less of jagged edges and make image less abstract.

All preview images are made in Forge, DPM++2M, Beta in SDXL resolution then upscaled 1.5 using DATx4 and Mixture of Diffusers. Some of the prompts are expanded using LLM.