MagicLab’s AI-powered humanoid runs outdoors for four minutes in new footage
MagicLab has unveiled new footage of its AI-powered humanoid running outdoors for four minutes. The impressive display showcases the robot's capabilities and agility in an outdoor setting. The humanoid's smooth movements and speed demonstrate the advancement of AI technology in robotics. MagicLab's latest development highlights the potential for AI-powered robots in various real-world applications.

Humanoid Robots by MagicLab
Chinese robotic firm MagicLab has released a video that shows its humanoid robot running outdoors for four minutes straight. The new demonstration comes after the company recently showcased its humanoid MagicBot’s improved walking mechanics, featuring a more human-like, straight-legged gait. Reports suggest the robot is gearing up for an upcoming half marathon in Beijing, where bipedal robots will race alongside 12,000 human participants.
Deployment in Factories
In December 2024, MagicLab deployed humanoid robots for job training in a factory. A video showed MagicBots handling tasks like inspection, material handling, picking, placing, scanning, and warehousing.
Advanced Technology
Founded in December 2023, MagicLab, also known as Magic Atom Robotics Technology, aims to transform industries such as search and rescue, logistics, and manufacturing with its advanced humanoid robots. With its full-scenario environmental sensing, the company’s third-generation MagicBot humanoid can easily navigate complex situations.
It can collect environmental data for navigation thanks to its LiDAR, head and waist RGBD cameras, three fisheye cameras, and ultrasonic sensors. MagicLab has created a navigation system that enables its robots to adjust to difficult terrain and changing circumstances. With a total carrying capacity of 88 pounds (40 kilograms), it has a 42-degree-of-freedom design and arms that can lift up to 44 pounds (20 kilograms).
Capabilities and Features
MagicBot boasts a battery life of up to 5 hours during continuous walking, operations, and other activities. The company claims that the robot can self-learn and improve its walking posture thanks to its motion control network. They can easily navigate uneven terrain because of their powerful anti-interference and anti-shock properties. It is appropriate for a range of industrial and operational scenarios due to its high task precision, which is achieved through the use of 6D visual servo technology and full-body imitation learning.
Precision and Dexterity
The remotely controlled MagicBot demonstrated remarkable dexterity in earlier demonstrations by folding baby clothing precisely, putting them in a basket, and skewering and cooking marshmallows. It displayed accuracy by dancing to upbeat music, watering a plant softly, and even performing a magic trick that turned an egg into a ball, demonstrating its versatility in managing small things and carrying out difficult tasks with ease. These features, which improve hand dexterity, are made possible by sophisticated micro high-torque servo actuators and multi-dimensional pressure sensors.
Object Manipulation
The developers stress that accurate object manipulation is made possible by six miniature servo actuators, guaranteeing responsive and easy control. Soft, slick, or flexible objects are among the many shapes, sizes, and materials that the robotic hand is adept at grabbing. According to the firm, with sub-millimeter positioning accuracy and the ability to handle several pounds, it sets a new standard for fine motor control in robotics.
Commercial Deployment
Earlier, MagicLab deployed a fleet of MagicBots for job training in a factory. A video shows MagicBots performing tasks like product inspection, material handling, parts picking and placing, scanning, and warehousing. Additionally, the robots exhibit small-scale teamwork, demonstrating their capacity for productive collaboration. To speed up model training, Magic Atom intends to enhance its AI data engine with automation and batch processing. The new MagicBot will be used in the commercial and industrial sectors after going into small-scale manufacturing in Q1 2025.