MixedUp 3D Blend 🎨 [Experiment]

This LoRA is trained on a mixed dataset focusing mostly on video game screenshots with ray-tracing as well some ue5 projects screenshots, I wanted to isolate what made them pop inside this LoRA.

Usage Tips :

• I personally use it at the weight of 0.9 so i recommend that

• the LoRA needs to have its trigger words "nv-vg3ds, detailed, realistic," in the prompt

  • I your character loses depth or still have anime face then add   (3d:0.5), rather then upping the lora weight

• you might need to used adetailer or inpaint for fix far away faces

• so far only tested on WAI-NSFW-illustrious-SDXL V14 (used in the showcase pics)

Observations  :

• while the result fell a little bit short of my expectation when i comes to character is somehow exceeded them when it comes to scenes especially urban ones, i think it might cuz of the cyberpunk2077 screenshots i added later on.

Feel free to rate this resource if you like it 💚