InSpatio-World 1.5 turns any image into an explorable 4D world

InSpatio-World 1.5 is an open-source spatial world model that turns static visuals and footage into explorable 4D scenes. Give it an image and a camera trajectory, and it will generate a video walking through that space along your exact path.

While version 1.0 relied entirely on reference videos, 1.5 expands the toolkit to almost any visual asset you have lying around, making controlled camera moves accessible without a 3D rendering pipeline.

What it can do

InSpatio-World 1.5 basically acts like a virtual drone pilot that understands spatial geometry.

Feed it a single image, and it will push past the original edges of the frame to infer hidden rooms, corridors, or landscapes while preserving visual style and spatial layout.

Feed it a set of synchronized photos, and you can loop a camera path around your subject to pull off seamless Matrix-style bullet-time sequences.

It also supports 360-degree panoramas for wide-range exploration, as well as full video clips. When given a dynamic video, the model keeps the original motion intact as the camera moves through the scene, letting you freeze time at any split second.

Under the hood, the system uses next-view prediction trained on dynamic real-world environments. It figures out what unseen angles should look like from existing viewpoints and locks subject identity and geometry across massive camera shifts.

Specs and getting started

The real selling point here is how light it is for everyday developers and creators.

The system is built on top of the Wan 2.1 framework using a 1.3B-parameter model. The entire download weighs less than 6GB, which fits comfortably inside most modern consumer GPUs.

Performance-wise, its 1.0 predecessor topped real-time dynamic benchmarks and hit 24 FPS on a single GPU. Version 1.5 continues that lightweight philosophy, though formal independent benchmark numbers are still rolling out.

The project code and deployment instructions are already live on GitHub and the official InSpatio project site. You can pull the repository and run the scripts locally.

Why it matters

We think this is a smart shift away from the typical prompt-and-pray video generators. The hardest part of AI cinematography has never been generating pixels—it has been getting the camera to move where a director actually wants it to go.

By taking direct camera paths from images, panoramas, and clips, InSpatio-World 1.5 becomes a practical tool for previz artists, indie animators, and spatial robotics researchers. Instead of re-rolling random text prompts, you can now block out cinematic camera moves on a single concept sketch.

Keep in mind that radical camera swings into completely obscured areas will still hit algorithmic hallucinations. But at under 6GB and capable of running on a home desktop, it is easily one of the most accessible spatial tools available right now.