Character LoRA trained on Flux.1 D to keep a consistent face and look for an original AI-generated character (not a real person). The dataset consists of AI-generated images in various angles, expressions, outfits, lighting and backgrounds. Use the trigger word "som_900" in the prompt. Recommended LoRA weight: 0.8-1.0.
Description
v1: first version. Flux.1 D, 10 epochs, network dim 16, trigger word som_900. Selected epoch: [10]


