AgiBot’s LinkCraft Turns Everyday Videos into No-Code Robot Performances

AgiBot’s LinkCraft Turns Everyday Videos into No-Code Robot Performances
Source: AgiBot
  • LinkCraft turns human motion videos into robot performances using AI-based motion capture and imitation learning.
  • The no-code platform supports voice, emotion, and multi-robot storytelling, and is now in public beta for AgiBot X2.

AgiBot has introduced LinkCraft, a no-code platform for programming humanoid robots. Designed for the AgiBot X2, the system lets users generate robot performances by uploading simple videos with no programming or special hardware required. It is already being used in retail, entertainment, education, and research settings.

Source: AgiBot

Users create a motion sequence by recording a video of a human in action, such as dance, martial arts, or everyday gestures. They then upload it to the platform, and the system uses AI-based motion capture to analyze the movement. LinkCraft applies retargeting algorithms and a cloud imitation learning framework to generate a set of instructions for the robot. All of this is performed without motion capture equipment or programming experience.

Users can also add voice, facial expressions, and emotions to the robot’s performance. LinkCraft supports audio uploads, online recording, and voice synthesis with customizable tone and voice type. The system synchronizes speech with gestures and expressions to create natural, human-like delivery. A timeline editor lets users combine actions, sounds, and expressions into continuous sequences, supporting narrative-driven storytelling and synchronized performances with multiple robots. The platform also includes a built-in library with over 180 action sets across 11 scenarios and 140 expression templates to speed up content creation.

The public beta of LinkCraft is now available. The AgiBot X2 robot is currently in mass production and delivery, with shipments expected to reach thousands in 2025, according to the company.


🌀 Tom’s Take:

LinkCraft shows just how powerful vision-language-action models can simplify high-effort tasks. This no-code approach is especially useful in retail or service roles where greetings, product demos, or scripted routines can now be customized and reused with ease.


Source: PR Newswire / AgiBot