Spectral Labs SGS-1
A generative-AI foundation model that turns text, sketches, or 3D meshes into editable parametric STEP CAD parts
What is it
Spectral Labs SGS-1 is a US generative-AI foundation model that converts text, sketches, or 3D meshes into editable parametric STEP-format CAD B-Rep parts for actual manufacturing. What it produces isn't just a display model but a parametric solid part engineers can keep modifying in CAD software.
What problem it solves
Generic AI-generated 3D is mostly mesh models, hard to edit and not meeting the precise geometry needs for manufacturing, so engineers often have to rebuild CAD from scratch to go to production. SGS-1 directly outputs STEP B-Rep parametric parts, filling the gap between "generative AI" and "manufacturable CAD," letting designers start from text or a sketch and then adjust within the existing engineering flow. It suits mechanical design engineers, product-development teams, and manufacturing staff wanting to speed the modeling process from concept to manufacturable part. As a foundation model, the actual generation quality and performance on complex parts still need confirmation at the engineering-validation stage per project.
Key Features
- Text to parametric CAD parts
- Sketch and 3D-mesh input
- Output in STEP B-Rep format
- Editable parametric solids
- For actual manufacturing use
- Generative-AI foundation model
Pros
- Outputs editable parametric CAD, not a dead mesh
- Bridges generative AI and the manufacturable engineering flow
- Supports multiple inputs: text, sketch, mesh
Cons
- Complex-part quality still needs engineering validation
- As a foundation model, practical details vary by project
Use Cases
- Engineers quickly building parametric parts starting from text
- Turning 3D scan meshes into editable CAD
- Speeding the flow from concept to manufacturable part
Editor's Note
Spits out editable STEP parts directly — a key step for generative AI entering engineering CAD.
FAQ
What format does SGS-1 output?
It outputs editable parametric STEP-format CAD B-Rep parts that can be further modified in CAD software.
How does it differ from ordinary AI 3D generation?
Most AI generates mesh models, whereas SGS-1 produces parametric solid parts meeting manufacturing needs, closer to production-ready CAD.
Can I use 3D scan data as input?
Yes — it supports text, sketches, and 3D meshes as inputs to generate parametric parts.
