Please Donate Buzz for FLUX Lora Training !

FLUX v1.0 :

Trained on FLUX dev with 70 photos of Evangeline Lilly with detailed GPT-4 captions. Tested on FLUX 1.dev (full) and FLUX fp8 and FLUX nf4 ! Use around strength 0.8-1.0. 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:Evangeline_Lilly_FLUX_v1-000061:1>

SDXL v3.0 :

Trained on Juggernaut X with 240 photos of Evangeline Lilly . Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3, AlbedoBaseV2! Use with keyword : "evxlilly" . Use around strength 1.0-1.1. CFG 5.0-7.0. Clipskip 1. Can be used for example as follows:

Positive : {Artstyle}, {Character and scene description}, evxlilly, <lora:evlilly_juggerX_xl_1_wocap_merger_57_99_144_03_015_055_04_107_06-evxlilly:1.05>

Negative : ugly, deformed, airbrushed, photoshop, rendered, (multiple people), child

SDXL v2.0 160mb :

Trained on 150 photos of Evangeline Lilly on SDXL 1.0 base with better Lora presets. Less than half the size of the previous version. Use with keyword : "evxlilly"! Lora strength with prompt should be around 1.0. Following prompt structure on Juggernaut v7, realismEngine 2.0 or Juggernautv9 lightning works well:

positive : {art style}, {character descrioption and action},evxlilly, <lora:evlilly_32_small_xl_5_standard_wo_cap-merger_24_65_045_055_and_19_30_65_05_02_03-evxlilly:1>

negative : photoshop, rendered, (multiple people), blurry, child

SDXL v1.0:

LoRa of Evangeline Lilly trained on 200 images of her with the SDXL base model. Recommended strength of 0.9-1.1 for ComfyUi Lora. The dataset is biased to her smiling so a negative prompt such as (smile:0.6) or so could help reduce this if unwanted.