A 3D generation API creates a 3D model, a textured mesh, from one or more images of an object, and on some models from a text prompt. You submit a job, poll until it completes, and download the asset in a format such as GLB, OBJ, FBX, USDZ or STL.
A 3D generation API turns a picture or a description into a 3D asset you can open in a 3D editor, a game engine or a web viewer. The output is a mesh, the surface of the object built from polygons, with textures or materials that give it color and finish. Some models produce PBR materials, the standard set of texture maps for physically based rendering, so the object looks right under different lighting.
The most common input is image-to-3D: one photo or render of an object. Some models also accept several views of the same object, such as the front, back and sides, so the model sees the parts a single photo hides. A text prompt can also describe the object, on models that accept one.
You send POST /v1/3d/generations with the model ID and the input: an image URL in image, several images in images, a prompt, or a combination, depending on the model. Generation runs as an asynchronous job. The API returns a job_id at once, and you poll GET /v1/jobs/{job_id} until the status is completed. The finished job lists the generated files, each with a download URL, a file name and, where the model provides it, the format, such as glb.
Teams use 3D generation for game props and characters, product models for online stores and AR previews, concept models for design reviews, and prototypes for 3D printing. When you compare models, look at how faithfully the shape and textures match your input, how clean the mesh is, which inputs and formats the model supports, and how long a generation takes. For humanoid characters, a model that can output a standard T-pose or A-pose makes rigging easier. Side-by-side model comparisons help narrow the list.
3D models bill per generated asset: a flat rate per request, with tiers that depend on settings such as resolution, precision and whether textures are generated. An untextured mesh can cost less than a textured one.
Browse the 3D models and upload an image in the Playground to see the result before you write code. To call a model, create an API key in the dashboard, submit the job to https://api.empiriolabs.ai/v1/3d/generations and poll the jobs endpoint. The 3D generation reference lists the request fields and the result format.
Pay as you go, or choose a plan with weekly allowances. Current rates are on the pricing page.
3 models, cheapest first where they share a billing unit. Each model's page has its full rate card, parameters and code.
Endpoints
/v1/3d/generationsPriced per request
| Model | Lowest rate | Applies to |
|---|---|---|
| TRELLIS.2 4B Microsoft | $0.25$0.025 per request | 512 asset |
| Hyper3D Gen2 Deemos | $0.80 per request | Per generation |
| Hitem3D 2.0 Hitem3D | $1.60 per request | 1536 white model |
Common formats are GLB, OBJ, FBX, USDZ and STL. Each model page lists the formats that model can export.
Yes. Image-to-3D models build a textured mesh from one image of an object. Models that accept several views of the same object can also use the sides a single photo does not show.
Yes, on models that accept a text prompt. You describe the object, and the model generates the mesh and its textures. Other models work from images only.
3D models bill per generated asset, with tiers that depend on settings such as resolution and texturing. On EmpirioLabs AI you can pay as you go or choose a plan with weekly allowances.
TRELLIS.2 4B lists the lowest rate per request of the 3 3D Generation models priced that way, followed by Hyper3D Gen2, Hitem3D 2.0. The price table on this page shows each model's lowest rate and what it applies to.
Check out our pricing or reach out if you want your own model deployed on our stack.