Stable Fast 3D
Turns a single image into a textured, UV-unwrapped 3D mesh in under a second.
About
Stable Fast 3D turns a single image into a textured, UV-unwrapped 3D mesh in under a second, building on the TripoSR architecture with rework aimed at game-engine-ready output. Beyond raw geometry it performs mesh quality optimization, generates textures with proper UV unwrapping rather than vertex colors, disentangles illumination from surface appearance, and predicts material parameters, so exported assets drop into rendering pipelines with less cleanup. Inference for a single image needs roughly 6 GB of VRAM with default options, and the repo includes experimental support for Windows and for Apple Silicon via the MPS backend plus a ComfyUI integration. Installation follows the usual clone and pip install requirements path on Python 3.8 or newer, though the model weights are gated on Hugging Face and require an account token to download. Stability AI distributes it under the Stability AI Community License: free for research, non-commercial, and commercial use while you and your affiliates stay under 1 million dollars in annual revenue, beyond which an enterprise license is required. The repository holds about 1.8k stars.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- 3D Model Generation
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Intermediate (3/5)
- License
- Stability AI Community License
- Minimum VRAM
- 6 GB
- Added
- Jul 29, 2026
Related Tools
CUDA-accelerated Gaussian splatting rasterization library with Python bindings from the Nerfstudio team.
Generates high-resolution textured 3D assets from images or text with separate shape and texture models.
Generates explorable, mesh-exportable 3D worlds from text or image prompts via panoramic proxies.
Roblox's research foundation model that generates 3D shapes from text prompts for game assets.
Reconstructs 3D shape, texture, and layout of objects from a single cluttered real-world photo.
Reconstructs dense 3D scenes from uncalibrated image collections without known camera poses.