# Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention

arXiv: https://arxiv.org/abs/2505.17412 · HTML: https://mlsanity.com/paper/2505.17412/

## Key facts

- Cited by 110 papers in the map (data through 14 Oct 2026).
- Builds on 44 of its 52 references that are in the map.
- First posted on arXiv 23 May 2025.

## Builds on (most cited first, 30 of 44)

- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning](https://mlsanity.com/paper/2307.08691/) (2023)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [Objaverse: A Universe of Annotated 3D Objects](https://mlsanity.com/paper/2212.08051/) (2022)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [Objaverse-XL: A Universe of 10M+ 3D Objects](https://mlsanity.com/paper/2307.05663/) (2023)
- [Structured 3D Latents for Scalable and Versatile 3D Generation](https://mlsanity.com/paper/2412.01506/) (2024)
- [LGM: Large Multi-View Gaussian Model for High-Resolution 3D Content Creation](https://mlsanity.com/paper/2402.05054/) (2024)
- [Wonder3D: Single Image to 3D using Cross-Domain Diffusion](https://mlsanity.com/paper/2310.15008/) (2023)
- [One-2-3-45: Any Single Image to 3D Mesh in 45 Seconds without Per-Shape Optimization](https://mlsanity.com/paper/2306.16928/) (2023)
- [Deep Marching Tetrahedra: a Hybrid Representation for High-Resolution 3D Shape Synthesis](https://mlsanity.com/paper/2111.04276/) (2021)
- [3DShape2VecSet: A 3D Shape Representation for Neural Fields and Generative Diffusion Models](https://mlsanity.com/paper/2301.11445/) (2023)
- [CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets](https://mlsanity.com/paper/2406.13897/) (2024)
- [InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models](https://mlsanity.com/paper/2404.07191/) (2024)
- [Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention](https://mlsanity.com/paper/2502.11089/) (2025)
- [Instant3D: Fast Text-to-3D with Sparse-View Generation and Large Reconstruction Model](https://mlsanity.com/paper/2311.06214/) (2023)
- [Hunyuan3D 2.0: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation](https://mlsanity.com/paper/2501.12202/) (2025)
- [GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting](https://mlsanity.com/paper/2404.19702/) (2024)
- [SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformers](https://mlsanity.com/paper/2410.10629/) (2024)
- [GRM: Large Gaussian Reconstruction Model for Efficient 3D Reconstruction and Generation](https://mlsanity.com/paper/2403.14621/) (2024)
- [One-2-3-45++: Fast Single Image to 3D Objects with Consistent Multi-View Generation and 3D Diffusion](https://mlsanity.com/paper/2311.07885/) (2023)
- [ImageDream: Image-Prompt Multi-view Diffusion for 3D Generation](https://mlsanity.com/paper/2312.02201/) (2023)
- [Uni3D: Exploring Unified 3D Representation at Scale](https://mlsanity.com/paper/2310.06773/) (2023)
- [Michelangelo: Conditional 3D Shape Generation based on Shape-Image-Text Aligned Latent Representation](https://mlsanity.com/paper/2306.17115/) (2023)
- [TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models](https://mlsanity.com/paper/2502.06608/) (2025)

## Built on by (most cited first, 30 of 110)

- [SAM 3D: 3Dfy Anything in Images](https://mlsanity.com/paper/2511.16624/) (2025)
- [Native and Compact Structured Latents for 3D Generation](https://mlsanity.com/paper/2512.14692/) (2025)
- [Hunyuan3D 2.1: From Images to High-Fidelity 3D Assets with Production-Ready PBR Material](https://mlsanity.com/paper/2506.15442/) (2025)
- [TexVerse: A Universe of 3D Objects with High-Resolution Textures](https://mlsanity.com/paper/2508.10868/) (2025)
- [VoxHammer: Training-Free Precise and Coherent 3D Editing in Native 3D Space](https://mlsanity.com/paper/2508.19247/) (2025)
- [Ultra3D: Efficient and High-Fidelity 3D Generation with Part Attention](https://mlsanity.com/paper/2507.17745/) (2025)
- [LATTICE: Democratize High-Fidelity 3D Generation at Scale](https://mlsanity.com/paper/2512.03052/) (2025)
- [ShapeR: Robust Conditional 3D Shape Generation from Casual Captures](https://mlsanity.com/paper/2601.11514/) (2026)
- [ShapeGen4D: Towards High Quality 4D Shape Generation from Videos](https://mlsanity.com/paper/2510.06208/) (2025)
- [UniLat3D: Geometry-Appearance Unified Latents for Single-Stage 3D Generation](https://mlsanity.com/paper/2509.25079/) (2025)
- [Seed3D 1.0: From Images to High-Fidelity Simulation-Ready 3D Assets](https://mlsanity.com/paper/2510.19944/) (2025)
- [FullPart: Generating each 3D Part at Full Resolution](https://mlsanity.com/paper/2510.26140/) (2025)
- [WorldGrow: Generating Infinite 3D World](https://mlsanity.com/paper/2510.21682/) (2025)
- [UltraShape 1.0: High-Fidelity 3D Shape Generation via Scalable Geometric Refinement](https://mlsanity.com/paper/2512.21185/) (2025)
- [Motion 3-to-4: 3D Motion Reconstruction for 4D Synthesis](https://mlsanity.com/paper/2601.14253/) (2026)
- [Pixal3D: Pixel-Aligned 3D Generation from Images](https://mlsanity.com/paper/2605.10922/) (2026)
- [QuadGPT: Native Quadrilateral Mesh Generation with Autoregressive Models](https://mlsanity.com/paper/2509.21420/) (2025)
- [HY3D-Bench: Generation of 3D Assets](https://mlsanity.com/paper/2602.03907/) (2026)
- [MeshRipple: Structured Autoregressive Generation of Artist-Meshes](https://mlsanity.com/paper/2512.07514/) (2025)
- [Are We Ready for RL in Text-to-3D Generation? A Progressive Investigation](https://mlsanity.com/paper/2512.10949/) (2025)
- [LATO: 3D Mesh Flow Matching with Structured TOpology Preserving LAtents](https://mlsanity.com/paper/2603.06357/) (2026)
- [Speed3R: Sparse Feed-forward 3D Reconstruction Models](https://mlsanity.com/paper/2603.08055/) (2026)
- [SegviGen: Repurposing 3D Generative Model for Part Segmentation](https://mlsanity.com/paper/2603.16869/) (2026)
- [LaS-Comp: Zero-shot 3D Completion with Latent-Spatial Consistency](https://mlsanity.com/paper/2602.18735/) (2026)
- [Part-X-MLLM: Part-aware 3D Multimodal Large Language Model](https://mlsanity.com/paper/2511.13647/) (2025)
- [LATO.2: Factorized 3D Mesh Generation with Vertex and Topology Flow](https://mlsanity.com/paper/2607.10623/) (2026)
- [ZeroScene: A Zero-Shot Framework for 3D Scene Generation from a Single Image and Controllable Texture Editing](https://mlsanity.com/paper/2509.23607/) (2025)
- [MoCA: Mixture-of-Components Attention for Scalable Compositional 3D Generation](https://mlsanity.com/paper/2512.07628/) (2025)
- [PEGAsus: 3D Personalization of Geometry and Appearance](https://mlsanity.com/paper/2602.08198/) (2026)
- [RecGen3D: Reconstruction-Guided 3D Generation in a Shared Canonical Space](https://mlsanity.com/paper/2604.01479/) (2026)

## Abstract (arXiv)

Generating high-resolution 3D shapes using volumetric representations such as Signed Distance Functions (SDFs) presents substantial computational and memory challenges. We introduce Direct3D-S2, a scalable 3D generation framework based on sparse volumes that achieves superior output quality with dramatically reduced training costs. Our key innovation is the Spatial Sparse Attention (SSA) mechanism, which greatly enhances the efficiency of Diffusion Transformer (DiT) computations on sparse volumetric data. SSA allows the model to effectively process large token sets within sparse volumes, substantially reducing computational overhead and achieving a 3.9x speedup in the forward pass and a 9.6x speedup in the backward pass. Our framework also includes a variational autoencoder (VAE) that maintains a consistent sparse volumetric format across input, latent, and output stages. Compared to previous methods with heterogeneous representations in 3D VAE, this unified design significantly improves training efficiency and stability. Our model is trained on public available datasets, and experiments demonstrate that Direct3D-S2 not only surpasses state-of-the-art methods in generation quality and efficiency, but also enables training at 1024 resolution using only 8 GPUs, a task typically requiring at least 32 GPUs for volumetric representations at 256 resolution, thus making gigascale 3D generation both practical and accessible. Project page: https://www.neural4d.com/research/direct3d-s2.