Model · 3D

TRELLIS

Microsoft's free, downloadable model that turns a picture (or, less well, a text prompt) into a textured 3D model you can save as a GLB file. MIT licence, so commercial use is fine, but you need a Linux PC with a 16 GB graphics card or a hosted demo.

Go to TRELLIS’s website

At a glance

Open or closed: Fully open
Weights and code downloadable under a standard open licence
Models and most code MIT; some submodules (e.g. diffoctreerast) have their own licences.
source · checked 12 Oct 2026 · checked on the maker's own page
Licence: MIT License (Permissive (e.g. Apache 2.0, MIT))
source · checked 12 Oct 2026 · checked on the maker's own page
Can I sell what I make?: Yes
MIT allows commercial use. Some optional submodules carry other licences; check them if you ship the code.
source · checked 12 Oct 2026 · checked on the maker's own page
Cost: Free option: Weights and code are free to download and run yourself.
Plans below.
How you use it: On your own computer, Through another service
README: NVIDIA GPU with at least 16 GB for the code overall (not per variant); tested only on Linux, Windows not fully tested. Free demo on Hugging Face Spaces. Outputs mesh (GLB), 3D Gaussians (PLY) and radiance fields.
source · checked 12 Oct 2026 · checked on the maker's own page
Skill needed: Technical
Needs Python, CUDA and a setup script; the Hugging Face demo needs no setup.
source · checked 12 Oct 2026 · checked on the maker's own page
Made by: Microsoft
Licence copyright: Microsoft Corporation. Country not stated on pages read.
source · checked 12 Oct 2026 · checked on the maker's own page
Status: Available
Still published. Microsoft has since released TRELLIS.2 (4B, image-to-3D with PBR textures, MIT) in a separate repo, github.com/microsoft/TRELLIS.2; it isn't in the catalogue yet.
source · checked 12 Oct 2026 · checked, but the page was unclear or partial

What you can make with it

A 3D object or scene, from a photo (main use)
A 3D object or scene, from a description
A 3D object or scene, by changing a 3D scene I have

image to 3D modeltext to 3D model

README recommends making an image first then using the image model; text models are 'less creative and detailed'. 'Local 3D editing' and asset variants are mentioned as capabilities.

source · checked 12 Oct 2026 · checked, but the page was unclear or partial

Plans and prices

No paid plans listed.

source · checked 12 Oct 2026 · checked on the maker's own page

Where you can run it

Hugging Face

Other options for “A 3D object or scene, from a photo”

3D AI Studio, ComfyUI, CSM (Cube), Dzine, fal, Hugging Face, Hunyuan 3D (Tencent), Hunyuan3D 2, Hyper3D Rodin, InstantMesh, Kaedim, KIRI Engine, Krea, Meshy, Mixar, Polycam, Scenario, Sloyd, Spline AI, Stable Fast 3D, Stable Point Aware 3D (SPAR3D), Step1X-3D, Tripo AI, TripoSR

About this entry

Facts last checked 12 Oct 2026 against the maker’s own pages (each fact links to its source). Entry status: draft. Plain-English summary written by Plainly.