InStyle: Instant Appearance Stylization of 3D Shapes

1University of Amsterdam 2Bosch Center for Artificial Intelligence

TLDR

  • We introduce InStyle, a diffusion-based feed-forward 3DGS style editor that delivers geometry-preserving, multi-view-consistent stylization in a single pass (~0.26 s), without per-asset test-time optimization, while generalizing to unseen assets and real-world captures.
  • We learn a structured geometry–appearance factorized latent space from spatially grouped Gaussians with controlled cross-branch sharing, and distill a frozen 2D editor through multi-view distribution matching for coherent 3D appearance editing.

Overview

Overview of our method. (1) Latent space learning: Given input Gaussians, our method first groups them based on spatial proximity and encodes them into group-structured disentangled latent spaces, with controlled cross-branch feature sharing. (2) Latent diffusion pre-training: A denoiser learns to denoise the noisy appearance latent conditioned on a text embedding, capturing 3D appearance priors. (3) Latent editing: Once 3D priors are captured, the denoiser is adapted into an editing function that maps the input appearance latent to a modified latent, guided by the geometry latent and trained via multi-view distribution-matching distillation from a frozen 2D editor. At inference, InStyle generates high-quality, 3D-consistent stylizations from text prompts in a single feed-forward pass, fully eliminating test-time optimization. Trainable models at each stage are denoted.

Results

Original

"Make it golden coloured."

Input Asset

"Make it in sunset palette"

Input Asset

"Make it Barbie style."

Input Asset

"Make it in cyberpunk style."

Input Asset

"Make its colours look like rainbow."

Input Asset

"Make it Starry Night Van Gogh painting style."

Interactive Scene Editing

With its ~0.26 second inference time, InStyle supports interactive scene editing ⚡

Qualitative Comparison

Real-world Generalization

InStyle generalizes to complex real-world 3DGS scenes by extracting target objects, stylizing them, and reinserting them into the original scene despite clutter and capture artifacts.

BibTeX

@article{ocal2025gaussianblender,
  title={GaussianBlender: Instant Stylization of 3D Gaussians with Disentangled Latent Spaces},
  author={Ocal, Melis and Xing, Xiaoyan and Li, Yue and Vien, Ngo Anh and Karaoglu, Sezer and Gevers, Theo},
  journal={arXiv preprint arXiv:2512.03683},
  year={2025}
}