TripoSR by stabilityai

TripoSR by stabilityai

Builds a 3D model from a single image in one forward pass. It was made by Stability AI and Tripo AI for developers, artists and researchers who want quick 3D from a picture.

Made by stabilityai on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: TripoSR is a 3D reconstruction model that takes a single image and produces a 3D model of what is in it, without a slow per-object optimization step. Stability AI and Tripo AI built it together. Its design follows the LRM network, with changes to the training data, the model and the training itself.

Who it suits: It suits developers, 3D artists and researchers who want a quick 3D starting point from one picture.

Industry | Who uses it | How they use it | Example
Games | 3D artists | Get a rough 3D shape from concept art or a photo | An artist feeds in a picture of a prop and takes the result into a 3D tool
Product | Design teams | Preview an object in 3D from a single image | A team turns one product photo into a 3D model to look at from other angles
Research | 3D vision researchers | Study single image reconstruction | A researcher compares its output with other reconstruction models

If you want faster generation or game-ready assets, you can skip it, since the card points to a newer model called SF3D.

Where to use it:

• The model's code repository on GitHub, which holds the usage instructions
• A Gradio demo hosted on Hugging Face
• Windows, macOS or Linux with Python, for downloading the files

Use case | Who | What they get back
Image to 3D model | Artist or developer | A 3D model rebuilt from one image
Quick shape preview | Designer | A rough 3D view of an object
Reconstruction research | Researcher | Outputs to compare with other methods

How to set it up:

• Python on your computer
• The Hugging Face tools, to download the model files
• The code repository for the model, since the card sends you there for usage steps

1. Install the Hugging Face hub package.
2. Download the model files.
3. Follow the usage steps in the code repository, which the card names as its guide. The card itself gives no run command.

pip install -U huggingface_hub
hf download stabilityai/TripoSR

How to use it: You give the model one picture and it returns a 3D model. The card sends you to its code repository for the exact steps, and also offers a Gradio demo you can try in a browser. The only command it shows is the download step.

hf download stabilityai/TripoSR

Tool | What it does
hf download | Fetches the model files to your computer
Gradio demo | Lets you try the model in a browser
Code repository | Holds the usage instructions the card refers to

Example: You upload a photo of a single object, such as a chair, to the Gradio demo. The model sends back a 3D version of the chair that you can look at from other angles.

For newcomers and experts: Most 3D tools need several views of an object. This model works from one picture. The easiest first try is the Gradio demo with a clear photo of one object on a plain background.

For experts, it is a feed-forward model, so it reconstructs in a single pass instead of fitting each object slowly. The training data are renders from a curated subset of the Objaverse dataset under the CC-BY license, made with a rendering method meant to look more like real photos. Training took 5 days on 22 GPU nodes with 8 A100 40GB GPUs each.

You provide one image of an object
The feed-forward network reads it once
It reconstructs the 3D shape
You get a 3D model back

Good to know:

• Input: a single image
• Newer option: the card points to SF3D for faster generation and more game-ready assets
• License: MIT
• Cost to run: none named
• Data it reaches: downloading uses the web and writes files to your disk, and the demo takes the image you upload
• Not for: deliberately producing or circulating 3D models meant to upset or insult viewers, or that reinforce stereotypes, as its misuse section says

On Hugging Face, 13 November 2026

• Task: image to 3d
• Downloads in the last 30 days: 168,894
• Likes: 689
• License: mit
• Quantized versions: 4
• Fine-tunes: 3
• Tags: 3d, image-to-3d

Links

• Model: https://huggingface.co/stabilityai/TripoSR
• Files: https://huggingface.co/stabilityai/TripoSR/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download stabilityai/TripoSR

Then follow the usage in its model card. Nothing in this kit runs by itself.