What is SpaceTimePilot?

SpaceTimePilot is an AI research project that uses video diffusion models to generate controllable videos from a single input, allowing independent changes to camera angles and time flow, such as slow motion or reverse playback.

Is SpaceTimePilot free to use?

Yes, it is a free open-source research project with code available on GitHub (release pending internal review), and no costs associated.

When was SpaceTimePilot released?

The project and paper were released on 31 December 2025, with demonstrations and code shared shortly after.

Can SpaceTimePilot generate videos with custom camera and time controls?

Yes, it excels at disentangling space and time, enabling features like bullet-time effects, arbitrary camera paths, and non-linear temporal patterns in generated videos.

SpaceTimePilot

Disentangle Space and Time for Controllable 4D Video Generation

From Adobe Research & University of Cambridge

Founder: Zhening Huang

Video & Animation

31 Dec 2025

1K+

Free

Pricing Model

Free

Starting Price

$0/Month

About This AI

SpaceTimePilot is a research-based AI video diffusion model that enables independent control over camera viewpoints and temporal dynamics from a single input video.

It allows generative rendering of dynamic scenes with arbitrary camera trajectories and temporal manipulations like bullet-time effects, slow motion, reverse playback, and complex space-time paths.

Image To Video Text To Video Video Editor Video Generator

Pricing

Pricing Model

Free

Starting Price

$0/Month

Key Features

Independent control of spatial camera motion and temporal scene dynamics
Supports bullet-time, slow motion, reverse playback, and zigzag time patterns
Generates continuous videos from monocular input with arbitrary trajectories
Uses temporal warping augmentation and custom Cam×Time dataset for training
Autoregressive inference for extended consistent video generation

Pros

Pioneering disentanglement of space and time in video diffusion models
Highly controllable outputs beyond standard video generation
Strong performance on both real-world and synthetic dynamic scenes
Enables creative effects like mixed space-time explorations
Open-source code and upcoming dataset release

Cons

Currently a research prototype with code under review for full release
Requires technical setup to run locally via GitHub repository
No hosted web demo or easy-to-use interface available yet
Computationally intensive as a diffusion model
Limited to research/academic use until broader release

SpaceTimePilot is an innovative research tool for advanced AI video creators and researchers interested in precise 4D control, offering unprecedented flexibility in manipulating space and time from single-video inputs.

FAQs

What is SpaceTimePilot?
SpaceTimePilot is an AI research project that uses video diffusion models to generate controllable videos from a single input, allowing independent changes to camera angles and time flow, such as slow motion or reverse playback.
Is SpaceTimePilot free to use?
Yes, it is a free open-source research project with code available on GitHub (release pending internal review), and no costs associated.
When was SpaceTimePilot released?
The project and paper were released on 31 December 2025, with demonstrations and code shared shortly after.
Can SpaceTimePilot generate videos with custom camera and time controls?
Yes, it excels at disentangling space and time, enabling features like bullet-time effects, arbitrary camera paths, and non-linear temporal patterns in generated videos.