Tech • AI • Robotics • Game
New Ameca Gen 3 AI Robot Is Getting Way Too Human
7/10TL;DR
Engineered Arts has launched Ameca Desktop Gen 3, a compact humanoid robot focused on facial expression, conversation and programmable AI behavior rather than walking or manual work.
KEY POINTS
A smaller Ameca built around the face
Ameca Desktop Gen 3 debuted on June 22, 2026 as the first third-generation Ameca. It measures about 21 inches tall, 16.5 inches wide and 10 inches deep, and weighs 19 pounds. Unlike the full-size 6-foot-2 Ameca, the desktop version drops the arms and legs and concentrates on the head and torso, where most human interaction happens.
Why the redesign centers on expression
The full-size Gen 2.6 robot from 2024 weighs 137 pounds and uses 61 motors, with 27 in the head alone. Those motors drive eyebrows, eyelids, lips and gaze, while none are in the legs. The design reflects Engineered Arts’ long-standing view that Ameca is a communication machine, not a labor robot, with much of its engineering budget devoted to expressions people instinctively read during conversation.
More sensors and a stronger voice
Gen 3 keeps the signature face with 27 facial motors and adds five in the neck for 32 total. It has HD eye cameras, a new wide-angle HD chest camera, a millimeter-wave radar sensor for short-range awareness, and a spatial audio microphone array that can locate where voices are coming from. Audio output is upgraded to a 40-watt speaker, double the 20-watt speaker listed for the full-size model.
Built for group interaction
A major new capability is parallel conversation, which allows the robot to separate and follow multiple voices even when people talk over one another. That addresses a common failure point for conventional voice assistants in noisy spaces. The system is aimed at reception desks, showrooms, classrooms and events where several people may engage the robot at once.
AI personality set by prompt
The robot runs Tridium, a three-layer software stack consisting of Tridium Robot OS, Tridium AI, and a web-based fleet management platform. Its new agent system lets users define personality, knowledge, language, voice and physical behaviors by writing a text prompt and uploading reference material. A company could, for example, create a branded showroom guide by describing tone and expertise in a few sentences.
Open to external AI tools
By default, the platform uses OpenAI ChatGPT for responses, Whisper for speech recognition and Amazon Polly for voice, but those components can be swapped out. It also supports direct integration with ElevenLabs, MCP connections to outside tools and data, plus REST and WebSocket APIs for developers. The result is a robot that can be connected to existing enterprise software and third-party AI services rather than operating as a closed device.
Latency remains a core challenge
Fast responses are still difficult in humanoid conversation. Earlier deployments reportedly saw answer times of about 2 seconds, while one independent study found some responses taking 6 to 15 seconds. To make delays feel natural, Ameca uses eye movements, glances, idle motions and other expressions so pauses resemble human thinking rather than system lag.
Designed to avoid the uncanny valley
Ameca’s gray, explicitly mechanical appearance is intentional. It is meant to signal clearly that it is a machine, reducing the discomfort often triggered by nearly human robots. The platform includes more than 50 pre-programmed expressions, plus tools to create custom combinations of lip sync, gaze, eyebrow movement, head tilts and subtle breathing-like motions.
Enterprise, education and healthcare are key targets
Engineered Arts is marketing the system for hospitality, exhibitions, education, research, entertainment and healthcare. Potential uses include reception, translation, teaching assistance, therapy support and medical training. In one training study, 67 percent of 12 participants said they would use Ameca as an extra way to practice patient interaction, though human actors were still seen as more lively.
Privacy, compliance and market position
The robot is certified for CE and FCC markets and is marketed with GDPR and CCPA compliance, local processing where possible, encryption, and enterprise security standards including ISO 27001 and SOC 2 Type 2. In late 2024, there were roughly 30 full-size Amecas in 27 countries and about 50 desktop units already in use. Engineered Arts does not publish pricing, but the full-size Ameca has been widely estimated at around $250,000, while the desktop model is sold through a direct purchase page.
CONCLUSION
Ameca Desktop Gen 3 sharpens the company’s core bet that the most valuable humanoid skills are eye contact, expression and conversation. Its success will depend less on locomotion than on whether its AI, timing and social behavior feel convincingly natural in real-world settings.
Ask a question