整理20个AI 3D工具链,展示从图片到游戏资产的全流程方案。
HOLY F*CK, Claude can now drive Blender. and that's just repo #13.
an entire AI 3D studio in 20 free repos.
Photo → 3D model → texture → rig → animate → ship.
01 TRELLIS → Turn one image into a 3D asset.
http://github.com/microsoft/TRELLIS
02 Hunyuan3D 2.1 → 3D assets with real PBR materials.
http://github.com/Tencent-Hunyuan/Hunyuan3D-2.1
03 TripoSR → Fast 3D from a single image.
http://github.com/VAST-AI-Research/TripoSR
04 Stable Fast 3D → Mesh with UV unwrap built in.
http://github.com/Stability-AI/stable-fast-3d
05 InstantMesh → Multiple views into one mesh.
http://github.com/TencentARC/InstantMesh
06 Shap-E → Text to 3D object.
http://github.com/openai/shap-e
07 Gaussian Splatting → Photos into a 3D scene.
http://github.com/graphdeco-inria/gaussian-splatting
08 gsplat → Splatting, GPU-fast.
http://github.com/nerfstudio-project/gsplat
09 nerfstudio → NeRFs made easy.
http://github.com/nerfstudio-project/nerfstudio
10 COLMAP → Photos into a point cloud.
http://github.com/colmap/colmap
11 Meshroom → Photogrammetry as nodes.
http://github.com/alicevision/Meshroom
12 UniRig → Auto-rig any 3D model.
http://github.com/VAST-AI-Research/UniRig
13 Blender MCP → Claude drives Blender for you.
http://github.com/ahujasid/mcp-for-blender
14 Blender → Model, rig, render.
http://github.com/blender/blender
15 Open3D → Point cloud toolkit.
http://github.com/isl-org/Open3D
16 PyTorch3D → Differentiable 3D for deep learning.
http://github.com/facebookresearch/pytorch3d
17 threestudio → Text-to-3D framework.
http://github.com/threestudio-project/threestudio
18 SAM 3D Objects → Any object in a photo, now 3D.
http://github.com/facebookresearch/sam-3d-objects
19 Three.js → 3D in the browser.
http://github.com/mrdoob/three.js
20 Godot → Ship it in a game.
http://github.com/godotengine/godot
you don't need a 3D artist anymore to get a first draft. you need a photo and this list.
Bookmark this. Go build a world ↓
来源:Replay · x.com