Galbot unveils RoboGesture for speech matched humanoid motion
Chinese robotics company Galbot has unveiled RoboGesture, a framework that generates humanoid robot gestures in real time based on spoken language. The system is intended to produce movements that reflect the meaning, rhythm, and tone of speech rather than relying on repetitive or disconnected gesture sequences.
According to Interesting Engineering, RoboGesture combines speech processing, artificial intelligence, motion generation, and robot safety controls. It was developed for physical humanoid robots by researchers from Galbot, Tsinghua University, Peking University, Beijing Institute of Technology, Harbin Institute of Technology, and Shanghai Qi Zhi Institute.
The researchers identify three technical problems behind the work: limited training data, the difficulty of mapping speech to suitable movement, and the safety risks involved when generated motions are transferred to physical hardware. Safety is particularly relevant for conversational humanoids, where arm and upper body gestures may occur close to people and must be produced without lengthy offline preparation.
RoboGesture is scheduled to be presented at the European Conference on Computer Vision in 2026. The available account does not provide quantitative evaluation results, identify a humanoid platform used for testing, or compare the framework with existing gesture generation methods. Those details will be needed to judge its latency, motion quality, and reliability on real hardware.
Source: Interesting Engineering
