FLUX v1.0 :

Trained on FLUX dev with 80 photos of Ana de Armas 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:Ana_de_Armas_FLUX_v1-000061:1>

v4.0 Juggernaut X :

Trained for Juggernaut X with 250 photos of Ana de Armas. Use with Juggernaut X, Juggernaut v7 or RealismEngine 2. Works best with Juggernaut X. Use with keyword : "anxarmas". Can be used for example as follows:

Positive : {Artstyle}, {Character and scene description}, anxarmas, <lora:anarmas_juggerX_xl_3_standard-anxarmas-000096:1>

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

v3.0 :

Trained on 250 more carefully selected images of Ana de Armas for 100 epochs with single repeat. The dataset captioning uses GPT-4 Vision. Recommended Lora strength is 0.8-1.2.

v2.1 :

Retrained on the same 420 images but with a better Lora model for more flexibility while attaining likeness. Keep strength between 0.9-1.1.

v2.0 :

Trained on 420 images of Ana de Armas for 100 epochs with single repeat and batch size 3. The dataset captioning is also done using a different method. Recommended ComfyUi Lora strength is 1.0-1.15.

v1.0 :

Lora trained on 40 images of Ana de Armas on the SDXL 1.0 base model. The recommended Lora strength for ComfyUi is 0.9-1.2.