LightAtlas aims to create beings made of light in the darkness. Whether this be a cosplay wearing a mesh of lights, or your soulmate made of starstuff, LightAtlas has flexibility for that variety. It has strong light/dark contrast by default. The main training trigger is LightAtlas, pair that with secondary words such as neon and wireframe, as per the example images, to get the full effect. I find that the phrase "LightAtlas style, a neon LightAtlas wireframe of" is particularly useful. Use LightAtlas further in your prompt to direct the lora, particularly if you have a long and complex prompt. For ZiT, strength of around 1-1.2 works just fine. Similar strengths can produce great results for Flux, but the more complex your prompt, the higher you'll want the strength, with 1.5 being more consistent. Krea-2 works well at a strength of 1. Lower strengths are more likely to give you a network of golden thread, than a glowing panoply of woven lights.

LightAtlas

LightAtlas Krea2.safetensors

Files with SHA256: 59b092b4aa528b5d37879ceabcd659cfaf87c05c0442c4dd48de5158f46225cc