InSpatio-World 1.5 — Beyond the Frame, Into the World

From Next-View Prediction to Spatial Intelligence

InSpatio-World 1.0 began with a simple idea: build a world beyond the original camera view from a single video. Today, we introduce InSpatio-World 1.5. It now accepts a single image, an image set, or a panorama, enabling much wider viewpoint changes while maintaining scene consistency. For cinematography, immersive experiences, and research in spatial and embodied intelligence, InSpatio-World 1.5 opens up new possibilities for developers and creators.

Real-Time Scene Roaming

Starting from a single image, explore a scene from a wide range of viewpoints.

1

Beyond the Original View

Move the camera into the scene and reveal areas hidden from the original viewpoint.

2

Realism and Style Consistency

Even under large viewpoint changes, newly revealed spaces remain believable and visually consistent with the source image.

3

Spatial Coherence

Maintain a coherent spatial structure through sweeping camera moves so every new view feels grounded in the same scene.

Demo Videos

Corridor

Castle

Bridge

Cloud Palace

Next-View Prediction in Dynamic Worlds

With video input, InSpatio-World 1.5 enables deeper exploration across space and time.

1

Spatiotemporal Exploration

Move the viewpoint freely through the scene and choose when to observe as the action unfolds.

2

Next-Viewpoint Prediction

Training the model to predict novel views from large-scale data captured in dynamic real-world scenes strengthens its spatial understanding and its ability to reason about how scenes evolve over time.

3

Embodied Intelligence

This capability supports cinematic camera work, viewpoint reasoning for spatial intelligence, and dynamic environment modeling for embodied agents.

Demo Videos

Rolling Oranges

Forest Boat Ride

Rainy-Night Cottage

Desert Tornado

Flour Dust

One Model for Multiple Input Types

From image sets to panoramas, InSpatio-World 1.5 turns varied source material into explorable spaces.

1

Flexible Inputs

The model accepts a single image, an image set, a panorama, or a video, enabling spatial exploration from a wide range of source material.

2

Bullet-Time Creation

Start with multiple synchronized images and design a camera path around the subject to create a smooth, coherent bullet-time sequence.

3

Generation Consistency

The model preserves subject identity, spatial structure, and scene details as the camera moves dramatically and time advances.

Demo Videos

Multi-image · Strawberry

Multi-image · Smartphone

Panorama · Celestial Palace

Panorama · Bookstore

Looking Ahead

Multiple input options let exploration begin with almost any source. Wider viewpoint changes expand what can be observed, while strong consistency supports continuous exploration across space and time. InSpatio-World 1.5 is now available to try, with its code released as open source. We invite researchers, developers, and creators to build on these capabilities and help shape the next generation of content creation, immersive experiences, and spatial intelligence.