Image-to-3D generator from Tsinghua and Tencent ARC Lab that builds meshes pixel-aligned to the input view.
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.
Pixal3D is a research image-to-3D generation model from Tsinghua University (BNRist), Tencent ARC Lab and Victoria University of Wellington. Instead of generating shapes in a canonical space and injecting image cues through attention, it generates 3D directly in a pixel-aligned way, consistent with the input view. The authors target pixel-level fidelity to the input image, which they describe as a central bottleneck of current 3D generators.
The method uses a pixel back-projection conditioner that lifts multi-scale image features into a 3D feature volume, a VAE that compresses pixel-aligned sparse SDF into sparse latents, and a two-stage generative process that predicts coarse structure and then detailed latents, decoded into a mesh. It extends to multi-view input by aggregating back-projected feature volumes, and a modular pipeline produces object-separated 3D scenes from images. The project page compares results against TRELLIS 2 and Hunyuan3D V3.1. A YouTube reviewer reported it beating both and put the total model size at about 24 GB.
The project page provided no code, weights or license links, and no paper link, though the citation gives arXiv:2605.10922 (2026). Release status and licensing should be checked against the project page and the arXiv paper.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source image-to-3D generator using pixel-aligned generation for accurate geometry and texture. Beats Hunyuan 3D and Trellis 2 in comparisons; models total 24 GB.
Closest neighbours by meaning, not just by category.
Model
Research model that unifies 3D scene generation and reconstruction in a single pixel-space diffusion model.
Model · Hi3D
Open-source image-to-3D and text-to-3D generation model from Hi3D, formerly Sparc3D, trained on ~2M curated 3D assets.
Model · Tencent
Tencent's unified multimodal model for 3D understanding, text-to-3D generation, instruction-based editing and part generation.
App · Leonardo AI
Leonardo.Ai feature that generates 3D models from a single image, using multi-angle reference views.
Model
Research model that builds complete 3D assets from sparse, unposed images by jointly doing reconstruction and generation.
Research
Research method that generates compositional 3D scenes from a single image via object-to-scene distillation.
