Meshy 6 Image to 3D turns a single reference image into a textured 3D mesh. You pass one image and receive a downloadable model, so a photograph or a concept render becomes a usable asset without manual sculpting or retopology.
The endpoint is OpenAI-compatible and is served at /v1/images/generations, so an existing image-generation client works by changing the model id. The response carries a url and a b64_json field for the asset, with usage reported in a meta object.
| Developer | Meshy |
|---|---|
| Model ID | meshy/v6/image-to-3d |
| Modalities | Image, Text → Image |
| Input formats | JPG, JPEG, PNG, AVIF, HEIF — URL or Base64 data URI |
| Capabilities | Standard, low-poly and smart-topology modes; triangle or quad topology; target polycount 100–300,000; PBR textures; symmetry control; A-pose and T-pose |
Three reconstruction modes
model_type selects standard, which builds a general-purpose mesh, lowpoly, which targets a reduced polygon budget, or smart-topology, which produces cleaner edge flow. The default is standard.
A polygon budget you choose
target_polycount accepts 100 to 300,000, defaulting to 30,000 for standard and 4,000 for smart-topology, which is capped at 15,000.
Quad or triangle topology
topology selects triangle, the default, or quad, so the mesh can be handed to a modelling pipeline that expects quads.
Texturing and PBR
should_texture is on by default, enable_pbr adds physically based maps, and texture_prompt accepts up to 600 characters to steer the surface.
Symmetry and pose control
symmetry_mode takes off, auto, which is the default, or on. pose_mode accepts a-pose or t-pose for characters that will be rigged.
Game-ready props
The low-poly mode with an explicit polygon budget produces assets that drop into an engine without a separate retopology pass.
Character bases for rigging
A-pose and T-pose output gives a predictable starting position for skeletons.
Product and catalogue visualisation
A single product photograph becomes a mesh that can be lit and turned in a viewer.
Concept to asset
Concept art becomes a blockout model early, before a modeller commits time to it.
Sources: https://docs.aimlapi.com/api-references/3d-generating-models/meshy/meshy-6-image-to-3d
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
import requests r = requests.post( "https://api.aimlapi.com/v1/images/generations", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "meshy/v6/image-to-3d", "prompt": "A studio photo of a red ceramic mug on a white table" }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/images/generations", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "meshy/v6/image-to-3d", "prompt": "A studio photo of a red ceramic mug on a white table" }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/images/generations \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"meshy/v6/image-to-3d","prompt":"A studio photo of a red ceramic mug on a white table"}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Output | |
Billed per generation. Each response reports the credits and the USD spent for that call in its meta object.
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Meshy 6 Image to 3D This page | Image generation | |||
| Image generation at scale | ||||
| Image generation at scale |
It builds a textured 3D mesh from a single reference image. You pass one image and get back a downloadable model, so a photograph or a concept render becomes a usable asset without manual sculpting or retopology.
JPG, JPEG, PNG, AVIF and HEIF. The image is passed as image_url, either as a URL or as a Base64 data URI.
target_polycount accepts 100 to 300,000. The default is 30,000 in standard mode and 4,000 in smart-topology mode, which is capped at 15,000.
Yes. topology takes triangle, the default, or quad, so the mesh can be handed to a pipeline that expects quads. Three reconstruction modes are available through model_type: standard, lowpoly and smart-topology.
Texturing is on by default through should_texture. enable_pbr adds physically based maps, and texture_prompt accepts up to 600 characters to steer the surface.
pose_mode accepts a-pose or t-pose, which gives a predictable starting position for characters that will be rigged. symmetry_mode takes off, auto (the default) or on.
It is billed per generation, at $1.04 per generation through AI/ML API. Each response reports the credits and the USD spent for that call in its meta object.
.svg){kind=small}
