trends.barnas.net

task · first seen 8 Oct 2026

image-to-3D generation

Also: image-to-3D, single-image 3D generation

Generating 3D assets such as meshes from a single input image.

Image-to-3D generation produces textured 3D assets, usually meshes, from a single photograph. The sources show a mature task in an incremental phase. Hi3D 3.0 (Twinkle3D) and NOFACE3D target fine geometric detail such as inscriptions and thin structures. SUPRA-Tex focuses on relightable texture alignment, Coarse3D on controllable sparse-structure stages, and Gate3DGen on cleaning degraded real-world inputs. Sprint3D profiles and speeds up an existing public pipeline. On the code side, there are official refinement weights for Hi3D, a compute-adaptive tokenization repo, and a local studio that wraps TRELLIS.2 for rigging and game export. None of this introduces a new task; it refines a widely studied one.

The work is shifting from showing that the task is possible to fidelity, speed and robustness on real-world inputs. That shift matters for game, 3D-printing and asset pipelines that want to use these models in production. The apparent rise in mentions reflects steady output on an established task, not a new concept.

Analyst note (established), 17 Nov 2026.

Mentions per week

0219 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 106 on 14 Nov 2026 · 7 mentions in 28 days · 7 independent groups · papers → code · first seen 5 weeks ago · acceleration +1.20

Earliest sources

Latest sources

Mentioned together with

diffusion transformerflow matchingsparse voxel representationphysically based rendering3D geometry refinementimage purificationsparse voxel generationdiffusion modelsTRELLISauto-rigging