Idol

Idol

Turns one image into a 3D object made of Gaussian splats, in a Gradio demo you run as a Hugging Face Space. For artists and researchers who want to try single-image 3D.

Made by speridlabs on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Idol is a demo from Speridlabs that builds a 3D object from a single image. The 3D result is made of Gaussian splats. It runs as a Gradio app on Hugging Face Spaces.

Who it suits: It suits people who want to see what one picture can become in 3D, and people who want a small example of how such a demo is built.

Industry | Who uses it | How they use it | Example
3D art and games | Artists and prototypers | Turn one reference picture into a 3D object | Trying a character sketch as a 3D figure
Computer vision | Researchers | Test single-image 3D on the example images and their saved settings | Comparing two presets on the same kind of picture
Software | App developers | Read the Space files to see how a Gradio front end wraps a model | Reusing the Space layout for their own demo

You can skip it if you need a finished mesh for a game engine, because the demo describes its output as Gaussian splats.

Where to use it: The Space runs in a web browser through its Gradio interface. You can also copy the Space and run your own.

Use case | Who | What they get back
One picture to 3D | Artists | A 3D object built from Gaussian splats
Trying the examples | Newcomers | Each example image run with the settings saved for it

How to set it up: The source shows no install command, so this page has no code block. The Space needs a Hugging Face token to load two image encoders.

• A Hugging Face account with access to the dinov3-vith16plus-pretrain-lvd1689m encoder from the facebook organization.
• The same account with access to the FLUX.2-dev encoder from black-forest-labs.
• A secret named HF_TOKEN on your copy of the Space.

1. Request access to both image encoders on your Hugging Face account.
2. Create an access token for that account.
3. Add the token to the Space as a secret named HF_TOKEN.
4. Open the Space once the secret is saved.

How to use it: Start with one of the example images. Each one is listed with the settings it runs with, so you can see what the demo does before you change anything.

Tool | What it does
app.py | The user interface of the Gradio demo
model.py | The model that makes the 3D object
examples/presets.json | Holds the example pictures and the settings each one runs under

Good to know:

• Without an HF_TOKEN from an account with access to both encoders, the Space cannot load them.
• The source gives no limits on image size, run time or export format.
• The source does not describe how the model works inside. Read model.py if you need that.
• The license is Apache 2.0.

On Hugging Face, 6 November 2026

• Likes: 161
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/speridlabs/idol
• Files: https://huggingface.co/spaces/speridlabs/idol/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.