Text-to-3D

What is text-to-3D?

You write "a Victorian-style wooden chair" and the AI produces a three-dimensional model you can spin around and inspect from any angle. That's text-to-3D: turning a written description directly into a 3D asset, skipping the long slog of manual modeling.

How is it different from text-to-image?

From "a picture" to "an object"
An image is flat and has one viewpoint; a 3D model is solid, and every angle has to hold up as you turn it.
The output has to be usable
It can't just "look right" — it needs a mesh and textures you can import into a game engine or modeling tool and keep editing.

How does it work, roughly?

Picture first, then "thicken" it
Many methods borrow text-to-image to produce reference views from several angles, then "stitch" those views into a solid model.
Newer techniques like neural radiance fields
Approaches like NeRF and 3D Gaussian splatting rebuild three-dimensional structure straight from multiple viewpoints, with ever-richer detail.

What can you do with it?

Game props, film assets, product displays for e-commerce, prototypes for 3D printing — work that used to take a professional modeler days can now start as a rough draft from one sentence. Text-to-3D is turning "3D creation" from a specialist skill into something anyone can try.

Bottom line: text-to-3D is "a sentence for a solid object" — you describe it, the AI models it, and it holds up from every angle.