Please Donate Buzz for FLUX Lora Training !
FLUX v1.0 :
I will post an improved version soon, but since I already spent 3000 Buzz on this I might as well post it.
Trained on FLUX.1 [dev] with 175 photos of Emma Watson with detailed GPT-4 captions. Tested on FLUX 1.dev (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:Emma_Watson-000052:1>
v3.0 :
Trained on Juggernaut X with 175 photos of Emma Watson with an improved Lora training method for better likeness and flexibility. Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3 and AlbedoBaseV2! Use with keyword : "emxwtsn" . Use around strength 1.0. CFG 5.0-7.0. Clipskip 1. Can be used for example as follows:
Positive : {Artstyle}, {Character and scene description}, emxwtsn, <lora:emwatson_juggerX_xl_2_standard_wocap-emxwtsn-000089:1.0>
Negative : ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child
v2.0 :
Lora trained on 150 selected images of Emma Watson between 2010 and 2017 with more variety, captioned by GPT4-V and trained with an improved Lora model for improved flexibility and likeness. Recommended weight is again 0.9-1.1.
v1.0 and v1.1 :
Lora trained on 200 images of Emma Watson between 2010 and 2017 trained on the SDXL 1.0 base model. Recommended weight for ComfyUi is 0.9-1.1.
