People at Nvidia threw their stone with the GET3D project, a sort of DALL-E for 3D objects. In mid-May we ran into Shap-E, a similar model developed by the folks at OpenAI. But before creating three-dimensional models with prompts, what interests us most is being able to transform 2D images into 3D without hitches. That is the goal of the CSM platform.
The idea of describing objects so that an AI model generates them is very interesting, but even better would be to feed that same model with an image of the object to reproduce. Art, spare parts, rare pieces… the potential is gigantic. Slowly, algorithms are gaining precision, but none of that will help without ease of use, and that's where the CSM portal, Common Sense Machines comes in.
CSM: Turn 2D Images into 3D Models
All you need to use CSM is an account (as an alternative, you can use secondary credentials), a good-quality image of the object to convert (in JPG, PNG, or WEBP format), and a lot of patience. CSM will present a preview of its results, but when you click on "Generate 3D Mesh", you should prepare to wait several hours before getting the file. The format chosen by CSM is GLB, GL Transmission Format Binary file. Being a standard, you will have no trouble opening it with your favorite software.
Now, if you don't want to wait, CSM's gallery is packed with models ready for download. Statues, animals, trees, food, video game and movie characters... chances are you'll find something useful for your personal projects without suffering through long training sessions. To finish, all models generated under the free trial will be public and open source, while Enterprise mode ensures faster processing, custom training data, and privacy for models. Want API access? "Very soon".
Official site: Click here
