InSpatio-World 1.5

InSpatio-World 1.5

Lets you move a camera through a scene in real time with a 4D world model, using the keyboard in a browser demo. For people exploring interactive scene generation.

Made by blanchon 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: This Space is an interactive demo of InSpatio-World 1.5, a 4D world model. You pick an example, or add your own image or video, and then steer a camera through the scene with your keyboard. The video frames are generated as you move. It runs on ZeroGPU and uses a small re-implementation of the model.

Who it suits: It suits researchers and developers who want to feel how a world model responds to camera moves, and anyone curious about generated scenes that you can explore.

Industry | Who uses it | How they use it | Example
Research | Computer vision researchers | Steer through a generated scene to judge how well it holds together | Move forward and look around an example scene
Software | Graphics and AI developers | Read how the speed-ups are applied in a live demo | Study the cached compile step in aoti.py
Creative work | Visual artists | Upload an image and explore it as a moving scene | Fly through a still picture
Games | Prototype designers | Test camera control in a generated world | Try the keyboard controls on a sample

You can skip it if you need exact, repeatable renders, since frames are generated from your live key presses.

Where to use it:

• A web browser: open the Space page on Hugging Face, where it runs on ZeroGPU.
• Windows, macOS or Linux: run a copy with Python 3.12.

Use case | Who | What they get back
Explore an example scene | Anyone | A scene you can fly through with the keyboard
Turn your own image or video into a scene | Creators | Depth and cameras worked out for you
Study the code | Developers | A demo with a custom viewer and compile cache

How to set it up: The hosted page needs nothing from you. A local copy needs these things first. The repository does not say what hardware a local copy needs.

• Python 3.12 on Windows, macOS or Linux.
• pip, which installs the packages.

1. Install the huggingface_hub package with pip.
2. Download the Space's files into a folder with the hf command.
3. Move into that folder.
4. Install Gradio 6.28.0, then the packages listed in requirements.txt.
5. Start the app with Python.

pip install -U huggingface_hub
hf download blanchon/inspatio-world-v1.5 --repo-type space --local-dir inspatio-world-v1.5
cd inspatio-world-v1.5
pip install gradio==6.28.0
pip install -r requirements.txt
python app.py

How to use it: Pick an example, or add your own image or video, in which case a depth model works out the depth and the cameras. Press Start and click the viewer so it takes your keys. Then move and look around with the keyboard.

Tool | What it does
Start | Begins the real-time run
W, A, S, D | Move through the scene
Arrow keys, or Q, E, R and F | Look around
Space and Shift | Go up and down

Example: You choose an example scene and press Start. You click the viewer, hold W to move forward and tap the arrow keys to look left. The frames follow your keys, with a short lag of under a second.

For newcomers and experts: If you are new, pick one of the examples and press Start. Click the viewer, then try W and the arrow keys. Upload your own picture only after that works.

If you know the field, frames are generated in groups of 12 from the camera path your keys make, about 0.8 seconds behind. Speed comes from FP8 precision via torchao in the DiT and one Inductor-compiled block that serves all 30 blocks and is stored on the Hub. The TAEHV decoder, cached text values and a depth-splat renderer help too. The viewer is a custom Svelte component.

You pick an example or add media
Depth-Anything-3 works out depth and cameras
Your keys make a camera path
Frames are generated in groups of 12

Good to know: Frames arrive roughly 0.8 seconds behind your keys, because each group of 12 is made from the path so far.

• License: apache-2
• Cost to run: none named
• Data it reaches: an image or video you add, which is processed to get depth and cameras
• Not for: exact, repeatable renders, since the frames follow your live key presses

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/blanchon/inspatio-world-v1.5
• Files: https://huggingface.co/spaces/blanchon/inspatio-world-v1.5/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.