The version has well inherited the detail advantages of Flux, and made the background more in line with the urban environment of domestic cities, rather than the scene bias of some parts of East Asia.
Trigger word: xiancheng
Weight: 0.7-1.2
On this basis, the performance of the characters has been adjusted. It is no longer the dark and oily East Asian faces, and both men and women can achieve good results. In order to maintain authenticity, the characters in this version are not of the internet celebrity or infantile aesthetic style, but the faces are not locked, please replace the face mold for raw pictures.
It is worth mentioning that because this lora uses layered training, there is almost no training on the limbs, and the integrity of all limbs is close to the original version of Flux.
Comparison of LoRA effects
Description
Derivative version of East Asian humanities, trained using different training sets, not only retains the restoration of the rough and real scenes of East Asian towns and cities, but also comes with a retro filter effect, which is more suitable for the aesthetics of young people in East Asia than the Kodak and a host of other film filters, that is, the concept of both vintage and beauty.
Trigger word: none, can also fill xiancheng_cityWeight: 0.7-1.2Cue word structure: you can refer to the example chart, or fill in the words with the backpropagation tool.This version inherits the detailed advantages of Flux very well, and makes the background more in line with the East Asian urban environment rather than the scene bias of some East Asian regions.
And on top of that, the representation of the characters has been tweaked so that they are no longer black and oily East Asian faces, and both men and women are well realized. In order to maintain the authenticity, this version of the characters are not Netflix or juvenile aesthetic style, but there is no face lock, please replace the face mold for raw pictures by yourself.