Heads up — this is a small test release trained on a limited dataset. A full version trained on the larger dataset is currently cooking and will be published as a separate upload once ready. Expect broader style coverage and better consistency there.

Shiny CGI (SFW)

A stylized LoRA that pushes outputs toward a clean, cartoonish CGI look with amplified gloss on latex, vinyl, leather, and other reflective surfaces. Think stylized 3D animation with the specular highlights cranked up.

What it does

• Shifts the aesthetic toward smooth, stylized CGI — somewhere between Pixar/Arcane territory and modern 3D character art

• Boosts wet-look shine and specular reflections on latex bodysuits, catsuits, gloves, and vinyl clothing

• Keeps faces in a stylized-realistic range — polished, not photoreal, not flat cartoon

• Plays well with bold color palettes (reds, blacks, whites) and clean backgrounds

Best for

• Stylized character portraits and costume designs

• Superhero, performer, and fashion-editorial concepts

• Material studies where the shine is the point

• Getting a CGI feel without touching a full 3D pipeline

Recommended settings

• Strength: 0.7 – 1.0 (start at 0.8)

• Helpful prompt tags: shiny, latex, cgi, glossy, polished, reflective

• Resolution: native 1024×1024, upscales cleanly

• Pairs well with: strong material descriptors — e.g. "glossy red latex bodysuit", "polished vinyl surface"

Training details

• Base model: Z-Image Turbo (bf16)

• Network: dim 32 / alpha 32

• Learning rate: 1e-4, constant scheduler

• Optimizer: Adafactor

• Epochs: 30  ·  Repeats: 5  ·  Resolution: 1024

• Noise offset: 0.05

• Text encoder: frozen (not trained)

Content notice

SFW only. Trained exclusively on safe-for-work imagery — PG-13 ceiling. This is a style LoRA, not a lewd generator. It won't steer outputs toward NSFW content on its own, and that's by design.

Tips

• Neutral, green, or gradient backgrounds make the gloss pop

• Drop strength to 0.5 – 0.7 if the plastic look gets too strong for realistic scenes

• Stack with character LoRAs for stylized OC portraits — it layers cleanly as a style pass