Unverified50% confidenceFactExact time
腾讯混元3D系列采用多阶段生成范式：先通过图像生成多视图，再经由重建网络得到网格，最后进行纹理烘焙
1
Sources
50%
Confidence
Long-term
Relevance
7/13/2026
First Seen
Sources
开源3D生成模型实测：Trellis 2能否挑战付费产品
Related Claims
UnverifiedThe breakthrough of generative AI in the 2D image domain is extending toward 3D content generation, with mainstream technical paths including NeRF-based implicit 3D representation, diffusion model-based point cloud generation, and methods that reconstruct 3D meshes from 2D multi-view images.79% similarUnverifiedSingle-image 3D reconstruction models that emerged in 2023–2024 (such as OpenAI's Shap-E, Stability AI's TripoSR, and Tripo3D) are already capable of inferring complete three-dimensional geometric structures and texture maps from a single image.75% similarUnverifiedImages 2.0 has achieved qualitative breakthroughs in three dimensions: text rendering, detail fidelity, and multi-turn conversational instruction following.74% similarUnverifiedUltraShape 1.0采用端到端的单阶段生成架构，直接从输入图像预测完整的3D资产，省去中间的多视图生成步骤74% similarUnverified3D高斯溅射（3DGS）以显式高斯基元替代隐式网络，在保持重建质量的同时大幅提升渲染速度73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/502056API
curl https://kongchang.com/api/v1/knowledge/claims/502056MCP
get_claim(id=502056)