Skip to content
← Models
Open weightsAcademic research (Long et al.)/10 trustVerified

Wonder3D

Single-image 3D generation via cross-domain diffusion: consistent multi-view normals and color fused into a detailed textured mesh.

Visit official page →

About

Wonder3D tackles the hardest part of single-image 3D generation — keeping every view consistent. It fine-tunes a diffusion model to jointly generate multi-view normal maps and their matching color images, then fuses those views into a textured mesh with a geometry-aware normal fusion step. The whole thing takes about two to three minutes per image, and because it was built on strong 2D diffusion priors, it generalizes well to stylized and AI-generated input images, not just photos.

It landed as a CVPR 2024 Highlight and became one of the reference methods the next wave of image-to-3D papers compare against. The code and checkpoints are open, and the community has since wrapped it in Hugging Face demos and Colab notebooks. The catch is focal distortion: the authors themselves note it can struggle with real captured photos taken at extreme focal lengths, since the model assumes its own camera setup.

Specs

Context window

Not applicable — vision model; generates 6 fixed views of the input image

Pricing

Free / open weights

Free quota

Fully open — no API

API access

No official API — self-host from the xxlong0/Wonder3D repo, or try community Hugging Face demos

Weights

Open weights

Released

Not disclosed

Benchmarks

  • Paper (arXiv 2310.15008): qualitative and quantitative comparisons against SyncDreamer and Zero123-family baselines
  • CVPR 2024 Highlight; no headline benchmark numbers independently verified

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

image-to-3d3d-generationmesh-generationcomputer-vision

Trust score — 8/10

/10 trustVerified

Scored 8/10 (TIER verified, metrics-v1). Evidence: CVPR 2024 Highlight paper (arXiv 2310.15008) with comparisons against SyncDreamer and Zero123-family baselines; open code and weights on GitHub (xxlong0/Wonder3D) with strong community uptake; corroborated by Hugging Face community demos. Strengths: peer-reviewed, open weights, widely reproduced. Docked for: (1) the repo's English README claims MIT while its Chinese README claims AGPL-3.0 — licensing is genuinely ambiguous, verify before commercial use; (2) headline claims are author-reported on academic comparisons. License: ambiguous (MIT vs AGPL-3.0).

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
8.0/10
Maintenance · recency, releases, issue load
6.0/10
Openness · license + docs
8.0/10
Manual review · sanity read + red-flag check
8.0/10
  • · CVPR 2024 Highlight paper (arXiv 2310.15008) with comparisons against SyncDreamer and Zero123-family baselines
  • · Open code and checkpoints on GitHub (xxlong0/Wonder3D), widely wrapped in HF demos and Colabs
  • · Project page live at xxlong.site/Wonder3D/
  • · License ambiguity: English README says MIT, Chinese README says AGPL-3.0 — verify before commercial use

Sources