LingBot-World (Robbyant)

Robbyant's open-source interactive world model that generates the next frame in real time from keyboard and camera commands.

LingBot-World is a world model Robbyant open-sourced in January 2026, adapted from the Tongyi Wanxiang Wan2.2 video-generation model. Given a starting image or text description, the user controls it step by step with keyboard actions or camera pose, and the model generates the corresponding footage — the paper calls this a world simulator. The company reports it keeps a scene consistent over minutes-long durations (called long-term memory), and interacts in real time at 16 frames per second with under 1 second of latency; code and weights are released under the Apache 2.0 license, aimed at narrowing the gap between open- and closed-source world models, for content creation, games, and robot learning. Version 2.0 (LingBot-World-Infinity), from July 2026, supports interaction of unlimited duration, with a distilled real-time version driving 720p, 60fps video, offered at 14B and 1.3B sizes.

ExampleGiven an indoor photo as a starting point, pressing W moves forward and A/D turns, and the model generates the following footage from the corresponding viewpoint, segment by segment.

Also called

LingBot-World 2.0, LingBot-World-Infinity, Advancing Open-source World Models

Related

World Model · Interactive World Model · Genie 3 · Wan (Alibaba Video Generation Model) · Autoregressive Video Generation · Robbyant

Sources

arXiv 2601.20540: Advancing Open-source World Models
GitHub: Robbyant/lingbot-world
arXiv 2607.07534: Infinite Worlds with Versatile Interactions (LingBot-World 2.0)

As of

2026-07