Idol Turbo

Idol Turbo

Turns one image into a 3D object about 12x faster than Idol, using a distilled version that needs 6 steps instead of 50. Made for people who want quick image-to-3D results.

Made by akhaliq 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 Turbo is a Gradio demo that makes a 3D object from one image. It is a distilled version of speridlabs/idol that takes 6 steps instead of 50, with no guidance pass, which cuts the network evaluations by 16.7x.

Who it suits: It suits 3D artists, game developers and researchers who want a fast image-to-3D result, and people who want to try a distilled model next to the original.

Industry | Who uses it | How they use it | Example
Game development | Indie developers | Turn a picture of a single object into a 3D draft | Test a prop image and view the 3D object
Design and 3D art | Artists | Get a quick 3D starting point from a reference image | Try several images in a row to compare drafts
AI research | Researchers | Compare a few-step model with the original Idol | Run the same image here and in the original Space

If you need the full 50-step output of the original model, you can skip it.

Where to use it:

• A web browser, on the Space's page on Hugging Face
• A local copy of the files, run on Python 3.12 under Windows, macOS or Linux
• A Space of your own, set to ZeroGPU hardware, if you copy it

Use case | Who | What they get back
Fast 3D draft | Artists and developers | A 3D object from one image, with far fewer steps
Model comparison | Researchers | A way to see how the 6-step weights compare with the original
Own copy | Developers | A Space of their own with the demo's UI and weights

How to set it up:

• A web browser, which is all the online page needs
• Python 3.12, for a local copy
• The Python packages listed in requirements.txt
• A Hugging Face token set as the HF_TOKEN secret, from an account that can use black-forest-labs/FLUX.2-dev, the source of the VAE encoder

1. Open the Space's page on Hugging Face to use it online. Nothing needs installing.
2. For a local copy, install the Hugging Face hub package and download the Space's files.
3. Move into the downloaded folder.
4. Install Gradio 5.49.1, then the packages from requirements.txt.
5. Start the app with python app.py.

pip install -U huggingface_hub
hf download akhaliq/idol-turbo --repo-type space --local-dir idol-turbo
cd idol-turbo
pip install gradio==5.49.1
pip install -r requirements.txt
python app.py

How to use it: Open the page, or the one your local copy starts, and give it an image. The app creates a 3D object from it. The page's layout comes from the original Idol demo by Speri Labs.

Example: Give the app a picture of one object, such as a chair. It runs the 6-step model and returns a 3D version of the object in the page.

For newcomers and experts: Open the Space page and try one picture of a single object. Nothing needs installing, and you see the 3D result in the same page.

The weights were distilled with a phase-LoRA method using 93 images of single objects, and guidance is built into them, so 6 steps without guidance can replace 50 guided ones. The repository holds app.py, model.py, turbo.py and an idol package. DINOv3 loads from a community mirror that was checked for matching numbers, so no Meta gate approval is needed.

You give the app one image
The 6-step Idol Turbo weights run on it
The VAE encoder and DINOv3 supply features
Gradio shows the 3D object in the page

Good to know: The model was distilled on single-object images, so single objects are the safe input. The online copy needs ZeroGPU hardware.

• License: Apache 2.0
• Cost to run: a Hugging Face token stored as HF_TOKEN, from an account that can use black-forest-labs/FLUX.2-dev, and ZeroGPU hardware for the online copy
• Data it reaches: the image you give it, and models loaded from Hugging Face, including akhaliq/idol-turbo and ZhengPeng7/BiRefNet

On Hugging Face, 13 November 2026

• Likes: 10
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/akhaliq/idol-turbo
• Files: https://huggingface.co/spaces/akhaliq/idol-turbo/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.