"RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives"
European Conference on Computer Vision, September 2026
Abstract
We present 'RenderFormer-V2', a unified learned transformer-based neural rendering model, complementary to modern physics-based rendering systems, that can handle diverse light-transport effects such as caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials without per-scene training or specialized code. RenderFormer-V2 models global light transport as a sequence-to-sequence transformation. Following its predecessor, RenderFormer-V2 also employs a two stage process: a view-independent stage that resolves intra-scene primitive to primitive transport, and a view-dependent stage that transforms the internal neural scene representation into image pixels. Different from RenderFormer, our model employs a novel combined windowed-attention and rendering-informed attention sink in the view-independent stage to improve scalability while maintaining render accuracy. To further improve versatility, RenderFormer-V2 supports heterogeneous scene primitives, including environment maps and participating media, and it employs a material encoding independent of the underlying surface reflectance model that encodes material appearance via a novel neural embedding. We demonstrate the versatility of RenderFormer-V2 on a variety of scenes and perform an extensive ablation of the improved attention mechanism.
Download
Supplementary Material
Bibtex
European Conference on Computer Vision, September 2026
AbstractWe present 'RenderFormer-V2', a unified learned transformer-based neural rendering model, complementary to modern physics-based rendering systems, that can handle diverse light-transport effects such as caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials without per-scene training or specialized code. RenderFormer-V2 models global light transport as a sequence-to-sequence transformation. Following its predecessor, RenderFormer-V2 also employs a two stage process: a view-independent stage that resolves intra-scene primitive to primitive transport, and a view-dependent stage that transforms the internal neural scene representation into image pixels. Different from RenderFormer, our model employs a novel combined windowed-attention and rendering-informed attention sink in the view-independent stage to improve scalability while maintaining render accuracy. To further improve versatility, RenderFormer-V2 supports heterogeneous scene primitives, including environment maps and participating media, and it employs a material encoding independent of the underlying surface reflectance model that encodes material appearance via a novel neural embedding. We demonstrate the versatility of RenderFormer-V2 on a variety of scenes and perform an extensive ablation of the improved attention mechanism.
Download
- Zeng_ECCV2026.pdf (11.81Mb)
Supplementary Material
- Document: Zeng_ECCV2026_supp.pdf (55.75Mb)
- ArXiv version (arxiv)
Bibtex
@conference{Zeng:2026:RF2,
author = {Zeng, Chong and Dong, Yue and Peers, Pieter and Zhang, Lvmin and Agrawala, Maneesh},
title = {RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives},
month = {September},
year = {2026},
booktitle = {European Conference on Computer Vision},
doi = {https://doi.org/10.1007/978-3-032-37281-9_30},
}
author = {Zeng, Chong and Dong, Yue and Peers, Pieter and Zhang, Lvmin and Agrawala, Maneesh},
title = {RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives},
month = {September},
year = {2026},
booktitle = {European Conference on Computer Vision},
doi = {https://doi.org/10.1007/978-3-032-37281-9_30},
}