Open-source image to 3D models
The open models that turn one image into a 3D mesh, and what decides if you can use them. Checked 2026-11-06.
- Tencent
Hunyuan3D 2.1
Tencent Hunyuan Community Licence- Commercial use
- Yes, outside the EU, UK and South Korea, and below 1 million monthly active users
- GPU memory
- 10 GB for shape, 29 GB with texture
- Output
- Mesh with PBR texture
- Microsoft
TRELLIS.2
MIT- GPU memory
- 24 GB (TRELLIS: 16 GB)
- Output
- Mesh with PBR materials, including transparency
- Stability AI and Tripo AI
TripoSR
MIT- GPU memory
- About 6 GB
- Output
- Mesh with vertex colours
- Tripo AI (VAST)
TripoSG
MIT- GPU memory
- 8 GB or more
- Output
- Mesh, shape only
- Stability AI
Stable Fast 3D
Stability AI Community Licence- Commercial use
- Yes, for organisations under $1M annual revenue
- GPU memory
- About 6 GB
- Output
- Mesh with UV texture and material values
- Tencent ARC
InstantMesh
Apache 2.0- GPU memory
- Not stated by the authors
- Output
- Textured mesh
Which one to try
- Best geometry with an MIT licence: TRELLIS.2, if you have a 24 GB GPU.
- Strong shape and texture, outside the EU, UK and South Korea: Hunyuan3D 2.1.
- A small GPU or a laptop-class card: TripoSR or Stable Fast 3D, both about 6 GB.
- Shape only, for printing: TripoSG, or Hunyuan3D's shape stage.
Training data
Nearly all of these models learned from Objaverse, the Allen Institute's open dataset of 800,000 3D objects. For models you can download and use today, see the free 3D model library.
No GPU, no setup?
2dto3D turns one product photo into a realistic 3D model at real size in about 5 minutes, with AR, an embed snippet and export to 7 formats. From $2 a model.
Your first model is $2, refunded in full within 14 days if you do not like it. Try it with a sample photo