Stable Fast 3D
stability.ai
- Free plan
- Windows
- macOS
- Linux
- Web
- API
What Stable Fast 3D is
Stable Fast 3D turns one image of an object into a textured, UV-unwrapped 3D mesh. Stability AI says the model is trained to generate a 3D model in under one second. It performs a delighting step and predicts roughness and metallic material parameters for the object. The expected input is 512 × 512 pixels, and the official code example saves the reconstructed asset as a GLB file. The repository includes ComfyUI custom nodes and a sample workflow, and API access is available. Documented inference options include CUDA, CPU, and experimental Mac MPS; Windows support is marked experimental. Default settings use about 6 GB of VRAM for one image. The Community License is free for research and non-commercial use, and allows commercial use at annual revenue of $1 million or less; above that, an Enterprise License is required. Accessing model files requires accepting the license and agreeing to share contact information. Stability AI recommends testing the model and adding safeguards suited to the intended use.
Who it suits
It may suit game and VR developers, as well as professionals in retail, architecture, design, and other graphics-intensive fields. Users should consider the input, hardware, access, and licensing requirements before choosing it.
Good to know, and what to watch
- Creates a textured, UV-unwrapped mesh from one image.
- Outputs GLB in the official code example.
- Includes ComfyUI nodes and an example workflow.
- Documents CUDA, CPU, and experimental Mac MPS inference.
- Expects an image sized 512 × 512 pixels.
- Default settings use about 6 GB of VRAM.
- Windows support is marked experimental.
- Commercial use above $1 million annual revenue requires an Enterprise License.
Plans
Every plan and priceStable Fast 3D API
Community License
The verdict
Stable Fast 3D is focused on converting a single object image into a 3D asset, with material prediction and GLB output. Check the license threshold, model-file access terms, and hardware requirements against your project before using it.
What the maker publishes
- Access requirement
- The model repository requires users to agree to share contact information and accept the license agreement to access model files. source
- Company
- Stability AI identifies itself as the maker of Stable Fast 3D, and a company announcement describes it as headquartered in London. source
- Generation speed
- The model page says it is trained to create a 3D model from one image in under one second. source
- Hardware
- The repository documents CUDA, experimental Mac MPS support, and CPU inference; Windows support is marked experimental. source
- Headquarters
- London, England, UK source
- Image-to-3D
- Stable Fast 3D takes one image of an object and generates a textured, UV-unwrapped 3D mesh asset. source
- Input requirement
- The model expects an input size of 512 × 512 pixels. source
- Integrations
- The repository provides custom nodes and an example workflow for ComfyUI. source
- Licensing limit
- Organizations or individuals with annual revenue above US $1 million must obtain an Enterprise License before commercial use of the model or its derivatives. source
- Materials and lighting
- The model performs a delighting step and predicts per-object roughness and metallic material parameters. source
- Output format
- The official code example saves the reconstructed 3D model as a GLB file. source
- Resource use
- The README says default settings use about 6 GB of VRAM for one image input. source
- Safety
- Stability AI recommends that developers test the model and add safeguards appropriate to their own use cases. source
- Support
- Stability AI directs users to submit a support request through its Knowledge Base. source
- Who it serves
- Stability AI describes the model as serving game and VR developers and professionals in retail, architecture, design, and other graphics-intensive fields. source
Checked against the maker's pages on 5 Oct 2026.