Expanded dataset by ~120 images. Should be a bit less oversaturated for photorealistic images now. Also trained this version for a bit longer than the initial one. Still no particular trigger word. All main showcase gallery images were still done just with base SDXL and this Lora.