Mesh from text or image
One prompt or one picture becomes a textured 3D mesh — OBJ, GLB, STL, USDZ, or FBX — through one asynchronous job.
Model family · 3D
Prompts in Chinese. Props in 3D.
Hunyuan3D is the 3D generation family from the team behind the Hunyuan foundation models: text or a single image in, textured mesh out — and its API is documented around Chinese prompts, a real differentiator. In MML ONE, the Rapid tier feeds meshes straight into the 3D stage and character studio.
By Tencent Hunyuan
What it's best at
The open line of 3D generation.
What the Hunyuan3D family does best, from its own repos and API docs — capability first, stated plainly.
Openly released lineage
The 2.0/2.1 line ships open weights and even training code — one of the most-starred open 3D generators on GitHub, with a PBR material pipeline in 2.1.
A higher hosted ceiling
The hosted API line reaches Hunyuan3D 3.x: multiview input, up to 1.5 million faces, smart low-poly retopology, quad meshes, and sketch-to-3D.
Straight onto the stage
MML ONE calls the official Tencent Cloud API and imports the resulting GLB directly into the 3D stage as props and into the character studio.
Chinese prompts, first-class
The API is documented around Chinese positive prompts — for Chinese-speaking crews this is native ground, not a translation layer.
In MML ONE
One Hunyuan tier, ready to route.
Exactly what our catalog serves today.
Hunyuan 3D RapidThe fast tier: text or a single image to a textured GLB mesh, on a minutes scale.
Meshes import into the 3D stage as props and into the character studio as character models. The in-app catalog is the live truth.
In the MML ONE flow
Where Hunyuan 3D earns its place.
Generated meshes land as props on the 3D stage and as models in the character studio — set dressing and characters without a modeling pass.
The honest part
Know before you route.
The Rapid tier we run defaults to no PBR materials and takes text or a single image only — the Pro features (multiview, low-poly retopology, million-face meshes) are a tier we don't expose yet.
Generation is asynchronous and lands in minutes, not seconds — and jobs can fail with internal service errors (we retry once automatically).
Text prompts run through the Chinese-documented API with a 200-character cap on Rapid, and the newest 3.x models are closed and API-only, unlike the open 2.x line.
Public Alpha
Bring one film into the graph.
Start with a premise, a screenplay, or a folder of references. We'll set up your provider keys and walk through the first scene with you.