Mesmer
Mesmer is a lifelike humanoid robot system developed by Engineered Arts, designed from 3D scans of real people to achieve highly realistic human appearance, facial expressions, and micro-movements. Each Mesmer robot can be built as a fictional character or a faithful recreation of a real person down to skin texture and fine hair. The system features 22 custom servo actuators in the head and neck, full-body 61-actuator articulation, and sensors including cameras, depth sensors, LiDAR, and microphones to perceive body language, age, and mood. Built on Engineered Arts' TinMan telepresence software, Mesmer can be remotely controlled and updated via the cloud.
Primary use case: Entertainment, exhibitions, interactive public engagement, keynote events, education, and research requiring highly realistic human likeness and expressive interaction.
Height
187 cm
Weight
62 kg
Specifications
Physical
Height
187 cm
Weight
62 kg
Locomotion
stationary
Commercial
Availability
commercial
Primary Use Case
Entertainment, exhibitions, interactive public engagement, keynote events, education, and research requiring highly realistic human likeness and expressive interaction.
Related news
RoboColosseum Launches Real-World Robot AI Benchmark With Unitree Support
The initiative offers more than 100 robots and 10,000 free evaluation hours, while initial G1 tests reveal how far robot policies remain from reliably completing tasks.
Tesla’s Optimus Data-Collection Effort Comes Into Focus After Scoble’s “Robot University” Claims
Tesla's own recruitment listings confirm it is hiring workers to wear motion-capture equipment and complete prescribed physical tasks to collect training data for Optimus, including daytime and overnight schedules in Austin and dual-shift operations at Fremont Factory and Milmont Academy. Technology commentator Robert Scoble's October 28 post adds unverified details — 24-pound suits with up to 40 cameras, 3D sensors and three shifts around the clock at Palo Alto — that the listings neither confirm nor refute. The article contrasts Tesla's controlled, employee-led collection effort with Figure's crowdsourced Index platform, which reported 69,943 weekly active users and uploads equivalent to 35 minutes of video per second, while cautioning that neither approach alone demonstrates robot performance.
How AGIBOT’s Physical Reasoning Dataset Connects Planning With Robot Actions
AGIBOT WORLD Release 4, the “Physical Reasoning” dataset, links high-level instructions and task structure to robot action trajectories, detailed manipulation descriptions and spatial annotations. Its successful, erroneous and adversarial examples span Easy, Medium and Hard levels and nine reasoning abilities, including counting, precondition reasoning, memory, long-horizon planning and subtask ordering. AGIBOT says it has used the data to train its next-generation foundation model, but no quantitative evaluation isolating the dataset’s impact has been published.