TripoSR AI is an open-source model for rapidly generating 3D meshes from a single 2D image, developed by Tripo AI and Stability AI.
What is TripoSR?
TripoSR is an open-source, state-of-the-art model for fast 3D reconstruction from a single image. It takes a NumPy array of an image as input and outputs a mesh file (GLB format). The model runs on local hardware with GPU support and is designed for quick iteration.
Key Features
- Fast 3D reconstruction — Generates a textured 3D mesh from one image in seconds.
- Open-source — Code and model weights are publicly available on GitHub.
- Customizable parameters — Adjust seed, number of inference steps (default 50), CFG scale, and target face count for mesh simplification.
- Texture application — Separate step to apply textures via the Apply Texture button.
- Export format — Download the final model as GLB (glTF Binary).
- Workflow integrations — Compatible with ComfyUI, Blender, and API via community tools and guides.
Who is it for?
- 3D artists and game developers — Quickly create base meshes for further refinement in Blender or other tools.
- E-commerce teams — Convert product photos into interactive 3D models for online stores to boost conversion.
- Architects and designers — Generate 3D representations from concept sketches or photos.
- Hobbyists and educators — Learn about AI-driven 3D reconstruction with an accessible, open-source tool.
What can you do with TripoSR?
- Generate 3D assets for games — Upload concept art and receive a mesh ready for texturing in Blender.
- Create product configurators — Turn single product images into rotatable 3D models for web use.
- Build architectural visualizations — From a photo of a building, get a 3D mesh for immersive walkthroughs.
How does TripoSR work?
- Upload a single image (best with clean subject isolation).
- Click Generate Shape to produce a 3D mesh.
- Optionally click Apply Texture to add textures.
- Download the GLB file. Adjust parameters like seed, inference steps, and CFG scale in the Generation Settings for fine-tuning.
Pros and cons
- Pros: Fast processing, open-source, customizable parameters, integrates with common workflows.
- Cons: Texture resolution may be limited; community scripts exist to enhance it. Requires GPU for local use.
FAQ
Is TripoSR free?
Yes, TripoSR is open-source and free to use locally. The web demo is also free.
What formats does TripoSR export?
The primary export format is GLB (glTF Binary).
Can I use TripoSR in ComfyUI?
Yes, there is a custom ComfyUI node. Start with the TripoSR ComfyUI Workflow Generator to get a JSON workflow and setup commands.
How do I improve texture resolution?
Community scripts are available on GitHub to modify color sampling code. Use the TripoSR GitHub Install Helper to set up for experimentation.
Can I run TripoSR offline?
Yes, download the model from GitHub and run locally with PyTorch and CUDA.
Alternatives
- SF3D — A multi-view reconstruction alternative; see the TripoSR vs SF3D comparison for differences.
