Overview
CSM (Common Sense Machines) is an AI platform that generates high-quality, production-ready 3D models and worlds from images, text, and sketches. Its Cube product lets creators produce meshes, animations, and Gaussian splats in a unified workspace for games, simulation, and design.
Our take
CSM converts images and text into 3D meshes quickly, and the output lands where most of this category lands: excellent for concepting, blocking out a scene, or a background prop, and requiring retopology before it goes anywhere that needs to animate or run at frame rate. The image-to-3D path is noticeably more reliable than text-to-3D, which is worth knowing before you judge it on a text prompt.
Worth it if
Concept artists and game developers who need a mesh to iterate against today rather than a production asset.
Look elsewhere if
Anyone expecting clean, game-ready topology straight out of the export — that step is still yours.
✨ Features
Image to 3D
Generates detailed 3D meshes from a single image or photo.
Text to 3D
Creates 3D models and assets directly from text prompts.
Production-Ready Meshes
Outputs clean, retopologized meshes suitable for games and simulation.
Unified 3D Workspace
Combines meshes, animations, and Gaussian splats in one editor.
⚖️ Pros & Cons
Pros
Cons
💰 Pricing
Tinkerer
$0
- 10 one-time credits
- Test the platform
- Creative Commons (CC BY 4.0) license
Maker
$20 /mo
- 100 credits per month
- High-res 3D meshes
- Premium features
- Fast dedicated servers
Creative Pro
$60 /mo
- 400 credits per month
- High-res 3D meshes
- Premium features
- Full creative workflows
Studio & Enterprise
$2000 /mo
- Per seat per month
- 13,333 credits per month
- High-res meshes & premium features
- Team/enterprise support
Categories
❓ Frequently Asked Questions
Is the output usable directly in Unity or Unreal?
It imports, but it usually needs work first. Generated meshes tend to be dense and irregularly topologised, which causes problems with deformation when rigged and with performance at scale. Plan for a retopology and bake pass on anything that will animate or appear in quantity; static background geometry is more forgiving.
Should I start from an image or a prompt?
An image, whenever you have one. Reconstruction from a reference removes most of the ambiguity the model would otherwise have to invent, and the result is far more predictable. Text-to-3D is best treated as an idea generator — useful when you do not yet know what you want, unreliable when you do.