Improved flux version of the model trained on Civitai's service using a single caption. Seems to reproduce the character quite accurately, however some description may be required. The base trigger is "queenbeemon", but I recommend using keywords like horns, antenna, single purple braid, bodysuit, green eyes, etc that help describe the character. It seems to work busy with 3d, anime, and cartoon style images, but you will need to lower the weight for photorealistic images. Seems to work best with a strength of 1.0 to 0.8.