I will make another version of this soon but since I already spent 3000 buzz I might as well post it.
FLUX v1.0 :
Test with 16/16 for network alpha/dimension and very large dataset and portrait style images. Works best with images around resolution 896x1152 px.
Trained on FLUX.1 [dev] with 200 photos of Sadie Sink with detailed GPT-4 captions. Tested on FLUX 1.D (full) and FLUX fp8 and FLUX nf4 ! Use around strength 0.8-1.2. No keywords needed! Distilled CFG around 1-4 and CFG 1.0 (without negative prompt). Clipskip 1. Can be used for example as follows:
Positive : {Artstyle, Character and scene description in usual FLUX fashion}, <lora:Sadie_Sink_FLUX_v1_r1-000052:1.0>
I will make another version of this soon but since I already spent 3000 buzz I might as well post it.
FLUX v1.0 :
Test with 16/16 for network alpha/dimension and very large dataset and portrait style images. Works best with images around resolution 896x1152 px.
Trained on FLUX.1 [dev] with 200 photos of Sadie Sink with detailed GPT-4 captions. Tested on FLUX 1.D (full) and FLUX fp8 and FLUX nf4 ! Use around strength 0.8-1.2. No keywords needed! Distilled CFG around 1-4 and CFG 1.0 (without negative prompt). Clipskip 1. Can be used for example as follows:
Positive : {Artstyle, Character and scene description in usual FLUX fashion}, <lora:Sadie_Sink_FLUX_v1_r1-000052:1.0>