Meshy released version 5 in July, and the update targets the gap that has kept generative 3D a novelty: usability. Where earlier text- and image-to-3D tools produced a single fused mesh with muddy surfaces and chaotic topology, Meshy 5 outputs assets with cleaner geometry, separated PBR material maps (base color, roughness, metallic, normal), and — the headline feature — automatic rigging that attaches a working skeleton to humanoid and creature models so they arrive ready to pose and animate.
Watch: Meshy AI: How To Create 3D Models from Text in Seconds (YouTube)
PBR maps are what make it usable in a real pipeline
A generated 3D model is only as useful as the materials on it. A single baked-in texture looks fine in the generator’s preview and falls apart the moment you drop it into a scene with different lighting. Meshy 5 outputs proper physically-based rendering maps, so the asset responds correctly to a scene’s lights and can be re-lit, recolored, and composited like anything a human modeled. For artists building AR filters, game prototypes, or virtual sets, that’s the line between “reference I have to rebuild” and “asset I can ship.”
Auto-rigging collapses the slowest step
Rigging — building the skeleton and weighting the mesh to it — is tedious, skilled work, and it’s the reason a beautiful static model is often days away from moving. Meshy 5’s automatic rigging attaches a standard skeleton to generated characters and creatures, so a prompt-to-poseable-character loop that used to require a rigging artist now runs in minutes. It won’t match a hand-rig for a hero character in a film, but for background crowds, rapid previz, and indie work, generating and animating on the same afternoon changes what a small team can attempt.
Where it fits
Meshy 5 doesn’t replace a modeler for precision work, and generated topology still needs cleanup for close-up hero assets. But it slots neatly alongside the other capture-and-generate approaches this site tracks — photogrammetry and Gaussian splatting for real objects, Meshy for imagined ones — as another road to filling a 3D scene without modeling every prop by hand.