Please Donate Buzz for FLUX Lora Training !

FLUX v1.1 :

Leave comments as to which FLUX version you prefer! It would help me!

Trained on FLUX.1 [dev] with 88 photos of Vanessa Merrell with detailed GPT-4 captions. Compared to version 1.1 this was trained slightly more and on CIVITAI instead of locally. 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:Vanmerrell_FLUX_v1-000061:1>

FLUX v1.0 :

Trained on FLUX.1 [dev] with 88 photos of Vanessa Merrell 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:vanmerrell_flux_1_standard-000047:1>

v3.0 :

Trained on Juggernaut X with 210 photos of Vanessa Merrell. Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3, AlbedoBaseV2 and SDXXXL 3! Use with keyword : "vanxmerrell" . 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}, vanxmerrell, <lora:vanmerrell_juggerX_xl_1_st_wocap_merger_83_139_05_05-vanxmerrell:1>

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

v2.0:

Retrained on 200 images and with an improved model to increase flexibility and likeness. Lora strength can be kept at 1.0. The model has also been tested with SDXL 1.0 Turbo, SDXL 1.0 Base and Juggernaut XL 3.0.

v1.0:

LoRa of YouTuber Vanessa Merrell trained on 30 images on SDXL base. The recommended strength for ComfyUi is 1.1.