Volumetric Temporal Texture for Smoke Stylization using Dynamic Radiance Fields


Dongqing Wang (École Polytechnique Fédérale de Lausanne (EPFL)), Ehsan Pajouheshgar (École Polytechnique Fédérale de Lausanne (EPFL)), Yitao Xu (École Polytechnique Fédérale de Lausanne (EPFL)), Tong Zhang (École Polytechnique Fédérale de Lausanne (EPFL)), Sabine Süsstrunk (École Polytechnique Fédérale de Lausanne (EPFL))
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

The exemplar-based stylization of dynamic 3D volume sequences remains challenging in computer graphics. The difficulty lies not only in maintaining the styling features concerning a reference image across multiviews and time frames, but also in preserving visually plausible motion as the original smoke simulation, within a reasonable time and computational resources. In this work, we introduce Volumetric Neural Cellular Automata (VNCA), a novel method that synthesizes volumetric temporal textures for smoke stylizations in real-time. Our method formats radiance fields atop the self-emerging Neural Cellular Automata (NCA) and generates a dynamic sequence of stylized color and density volumes. The synthesized temporal textures align with the input sequence's overall motion patterns while matching the style of a reference image. We use flow-guided supervision to align the texture's temporal motion with that of the perceived smoke sequence, reducing the training time by over an order of magnitude. We demonstrate that VNCA can be easily adapted for mesh stylization, akin to solid texture modeling, extending its application beyond dynamic volume simulations.

Citation

@inproceedings{Wang_2025_BMVC,
author    = {Dongqing Wang and Ehsan Pajouheshgar and Yitao Xu and Tong Zhang and Sabine Süsstrunk},
title     = {Volumetric Temporal Texture for Smoke Stylization using Dynamic Radiance Fields},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_18/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection