The model is used to provide an effect similar to light-painting photography for the generated image, (mainly characters, in the non-realistic model, mechs and vehicles seem to have good effects). The realistic model is a bit like the one released by social platforms ( Can you try to generate some to remove the camouflage?), there is a certain degree of reduction. In 2d and 2.5d models, the generated light will be more and more like a magic effect. Take what you need for your use case.
Suggestions for generation: I used Moyou, mjmix, reV, and awpainting to generate the examples. Mecha and underwater have been used in conjunction with my other lora. The specific parameters can be seen in the image generation data. High-definition restoration is required. The generation weight is recommended to be 0.7 Between -0.9, if it is used with other lora, it is recommended to adjust it below 0.8 according to the situation. There are not many types of illustrations temporarily generated, and I hope to return some pictures (mainly due to the limitation of brain hole)
In terms of creative overview, it is that the light paintings generated by mj are too magical, and it seems that station C has never seen this variety before, so let me fill it in
The materials used include a certain book and some pictures that I have tried to shoot in light painting collections, as well as mj generated pictures to supplement human faces
The main starting word is "light painting", other words can be played freely, it seems very suitable to play with controlnet