Please Donate Buzz for FLUX Lora Training !

FLUX v2.0 :

Trained on FLUX.1 [dev] with 180 photos with a larger Lora than the previous version of Isabela Merced (Dora, Instant Family, Last of Us, Superman). 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:Isabela_Merced_large_dataset_FLUX_v4_merger_214570_050401_merger_214570030205_merger_07_03:1.0>

FLUX v1.0 :

Trained on FLUX.1 [dev] with 100 photos of Isabela Merced. 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:ismerced_wocap_flux_1_standard-000026:1>

v4.0 :

Trained on Juggernaut X with 200 photos of Isabela Merced with an improved Lora training method for better likeness and flexibility. Tested on Juggernaut X, Juggernaut v7, RealismEngine 2, RealVisXL3, AlbedoBaseV2! Use with keyword : "isxmerced" . 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}, isxmerced, <lora:ismerced_juggerX_xl_2_wocap_merger_6_95_05_05-isxmerced:1>

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

v3.1:

Lora of Isabela Merced trained another 5760 steps.The likeness has slightly improved. Keep strength at 0.8-1.2.

v3.0:

Lora of Isabela Merced retrained on the same larger dataset but with an improved Lora model helping imporve flexibility and likeness. Keep strength at 1.0-1.1.

v2.0 :

Trained on a much larger dataset of 180 images of Isabela Merced. Keep lora strength for ComfyUi at 0.95-1.05.

v1.0 :

LoRa trained on 59 images of Isabela Merced with the SDXL 1.0 base model. The training was done using Google Colab kohya-ss gui. The recommended strength for the LoRa in ComfyUI is 1.1 and in A1111 is 0.7-0.9.