Tsinghua team reveals X1, an interactive video model you can touch

The line between virtual characters and the physical world just got a little blurrier.

Reaching through the camera

Tsinghua University's Xmax AI team has unveiled X1, a real-time interactive video model designed to bring digital creatures into your physical surroundings.

Instead of tapping on a screen or clutching dedicated controllers, X1 runs off a standard phone camera. Users can simply reach out their bare hands, and the model recognizes the gestures in real time, letting you touch and manipulate virtual characters on the fly.

The team claims X1 achieves millisecond-level latency alongside infinite-length video generation. A free private beta is now open for signups on social media.

Why it matters

Interactive generative video is the logical next step after passive video clips. Most current models generate static video offline; turning generation into a responsive loop makes it feel like an actual companion rather than a looping GIF.

Anchoring this experience to a mobile camera is a clever move because it bypasses specialized hardware hurdles entirely.

Still, sustaining infinite video generation at millisecond latency without visual glitches will be the real hurdle. It is worth keeping an eye on how the early beta handles real-world environments.