CivArchive
    Indies vintage illustration [ SDXL 1.0 / FLUX.D ] - v1.0 - FLUX
    NSFW
    Preview 1Preview 2Preview 3Preview 4Preview 5Preview 6Preview 7

    Home trained FLUX and SDXL Lora on a RTX 5070 ti graphics card.

    SDXL version 1.0: Trained with Image Dataset

    FLUX version 1.0: Trained with Image Dataset

    FLUX version 2.0: Fine-tuned Lora more refined oriental illustrations

    FLUX version 3.0: Fine-tuned Lora to add Illuminated manuscript borders and more refined oriental illustrations

    Description

    4 repeats, 20 epochs, 41 images = 3280 steps

    Learning Rate : 8e-4, Seed : 42, Guidance_Scale : 1, Repeat trains per image : 4, Max Train Epochs : 20, Save every N epochs : 4, Sample Image Every N Steps : 50, Resize dataset images : 1024

    Training time: 06:00:24