Ant Lingbo Open-Sources Embodied Video Foundation

机器之心 · wechat · 2026-07-09

Synced introduced LingBot-Video, open-sourced by Ant Lingbo: a video generation foundation model tailored for embodied AI, featuring a DiT video pre-training paradigm designed specifically for robotic scenarios.

The project integrates internet video data with robot manipulation, navigation, and first-person perspective data. It also incorporates an MoE video framework and a multi-dimensional reward system across its architecture, data, and training processes to enhance physical plausibility and task relevance.

Related event: Ant Group Open-Sources LingBot-Video for Embodied AI(26 posts)→

Original post →

More from Embodied

Embodied channel →