Ze-Qiang Lai

Individual token selection and metadata caching unlock sparse attention for diffusion transformers without sacrificing generation quality, achieving up to $2.32\times$ speedups.

High-fidelity 3D texture generation is now achievable without any 3D assets, by cleverly transforming 2D images into effective training data.

Local chunk-based modeling in ChunkVAE reduces memory usage and computation time while enhancing 3D reconstruction quality, outperforming traditional global latent approaches.

PolyFlow achieves rapid, high-quality mesh generation while outperforming traditional autoregressive models, thanks to its innovative continuous topology embedding.

Diffusion Transformers get a 2x speed boost without sacrificing quality thanks to a new routing mechanism that dynamically skips computations based on input sample sparsity.

Unlock controllable 3D asset generation with HY3D-Bench, a meticulously curated and augmented dataset of 375k objects featuring part-level decomposition.

Open-source 3D asset generation just leveled up: Hunyuan3D 2.0 delivers high-resolution, textured 3D models that outperform both open and closed-source alternatives.