← Blog

Text to 3D Is Live — Type a Prompt, Download a 3D Model

Forge is live on three.ws. Type a description — a crystal wizard with a glowing staff — and get back a textured, interactive 3D model you can orbit, inspect in AR, and download as a standard GLB file. It runs entirely in the browser at three.ws/forge. No install, no signup, no 3D experience required.

It also works from photos: upload one to four images of an object and Forge reconstructs it as a 3D model. Prompt in, model out — both directions.

Try Forge — type a prompt →

How it works

Text→3D runs a two-stage pipeline. First, your prompt is rendered into a reference image by a state-of-the-art image model. Then the image is lifted into 3D by TRELLIS, a 3D reconstruction model that produces full geometry with baked textures. The output is a GLB — the open standard that loads in Blender, Unity, Unreal, three.js, and every major 3D tool.

What you do with the model

A generated model is not a dead end on three.ws — it feeds the rest of the platform:

Agents can use it too

The same generation pipeline is exposed through the three.ws MCP server, so AI agents — Claude, ChatGPT, Cursor, or your own — can generate 3D models programmatically as a tool call. Your agent describes an object; it gets a GLB back.

Why this matters

3D content has always had the steepest creation curve on the internet: modeling, UV unwrapping, texturing, exporting. Text→3D collapses that into a sentence. For three.ws — the platform where every AI agent has a 3D body — it means the supply of bodies, props, and worlds is no longer gated on 3D artists. Anyone who can describe a thing can make one.

Try it


← All posts