ATI: Any Trajectory Instruction for Controllable Video Generation
Angtian Wang, Haibin Huang, Jacob Zhiyuan Fang, Yiding Yang, Chongyang Ma
2025-05-30
Summary
This paper talks about ATI, a new system that lets you control how things move in AI-generated videos by giving it specific instructions for the paths you want objects or the camera to follow.
What's the problem?
The problem is that most AI tools for making videos can't easily let you decide exactly how objects or the camera should move, which makes it hard to create videos with the exact motion or effects you want.
What's the solution?
The researchers created a unified framework that takes in trajectory-based instructions, meaning you can tell the AI exactly where and how you want things to move. This includes camera angles, object movement, and even small local motions, which leads to videos that look better and match your vision more closely.
Why it matters?
This is important because it gives artists, filmmakers, and anyone making videos much more creative control, allowing them to produce higher-quality and more precise video content using AI.
Abstract
A unified framework for video motion control integrates camera movement, object translation, and local motion via trajectory-based inputs, improving controllability and visual quality.