FLUX V1: Trained on 12 (512x512) images using civitai default settings.

FLUX V2: Trained on 48 images (1024x1024) on tensor.art

SDXL: Trained on 40 images (1024x1024) on civitai