Deemos turns text prompts into textured 3D models with Rodin Gen-1

Generating flat images from text is old news by now, but creating fully textured 3D objects from a prompt is still the frontier everyone wants to crack.

AI and metaverse development studio Deemos just officially unveiled Rodin Gen-1, a generative AI model that builds 3D models directly from text descriptions. The team calls the system revolutionary, showing off models with surprisingly complex shapes and PBR (physically based rendering) textures straight out of the prompt.

Turning prompts into textured 3D assets

Unlike many early text-to-3D experiments that output blobby monochrome meshes, Rodin Gen-1 focuses on visual completeness. The inclusion of PBR textures means lighting, reflections, and surface details look ready for a modern render pipeline right away.

There is a big catch, though. The developers at Deemos openly noted that every asset generated by Rodin "should be retopologized" before serious production use.

Why it matters

For solo game devs, 3D artists, and prototyping teams, building concept assets from scratch eats up massive amounts of time. A model that delivers complex shapes alongside surface textures can serve as a massive shortcut for concept validation and digital whiteboxing.

We think Deemos made a smart move by being brutally honest about the retopology requirement. AI-generated 3D meshes are notoriously messy and topology-blind, meaning you cannot easily animate them or drop them into an optimized game engine without manual cleanup. Instead of pitching Rodin as an overnight replacement for 3D artists, framing it as a rapid prototyping generator keeps expectations grounded and makes it genuinely useful for existing production pipelines.