Research text-to-3D method using block-wise diffusion over shape tokens, about 5x faster than autoregressive baselines.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Block3D is a text-to-3D generation framework from researchers at Zhejiang University, Monash University and the University of Adelaide. It partitions a 1,024-token discrete shape sequence into contiguous blocks. The blocks are generated autoregressively, and all tokens within the active block are denoised jointly and in parallel with diffusion.
A confidence-guided intra-block correction step revises low-confidence tokens before each block is finalized, and the completed sequence is passed through a frozen shape decoder. On a held-out set from TRELLIS-500K, the authors report mean end-to-end generation time falling from 25.71 s to 4.99 s, a 5.15× speedup over a fine-tuned autoregressive baseline, without loss of geometric fidelity.
The project page shows generated meshes and composed scenes. It is a research release accompanied by an arXiv preprint. A YouTube creator reports that code has been released, but no repository URL appeared in the provided material, so licence and weights status are unconfirmed.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Text-to-3D generator that splits objects into blocks of shape tokens, generating blocks autoregressively and tokens within each block in parallel via diffusion. Over 5x faster than standard autoregressive methods, about 5 seconds per model; code released.
Closest neighbours by meaning, not just by category.
Model · Stability AI
Research method adding typed 3D constraint meshes (hull, avoidance, touch) as control for text-to-3D generation on TRELLIS.2.
Research
Diffusion framework that learns text-to-3D human motion generation from 2D poses extracted from monocular videos.
Model · Princeton
Fully open 3B text-to-image diffusion model from Princeton with released weights, training code, and data pipeline.
Model
Research model generating 3D indoor scenes as sparse voxels, refined from 25 cm to 1.5625 cm by an autoregressive cascade.
Model
Roblox research model that generates text-to-3D objects as separate part meshes from a user-defined part list.
Model · NVIDIA
NVIDIA autoregressive diffusion model for real-time, text- and constraint-controlled 3D human motion generation.