Research record R-004Commercially available

Ameca

By Engineered Arts · United Kingdom / United States · announced December 2021

The world's most expressive robot face, sold and rented to venues worldwide; it talks brilliantly and does no chores at all.

Broader research scope

This is a research record, not a curated recommendation.

The main field guide selects 14 companion robots for deeper review. This entry belongs to a separate 53-record index that also covers industrial, research, and developer systems.

Research summary

What the record says.

Ameca is Engineered Arts' flagship humanoid and the most famous robot face on the internet. Its gray-skinned, deliberately non-human-colored head can produce startlingly lifelike micro-expressions: smirks, blinks, double-takes and eye contact that survive close inspection. Since a viral debut in December 2021 it has become the default robot for expos, museums and television.

Engineered Arts restructured as a US company in December 2024 with a $10 million Series A and now operates from Redwood Shores, California alongside its original UK base. Ameca is genuinely commercially available to businesses on a quote basis, with rentals for events; third-party estimates put full units in the low-to-mid six figures. A desktop line (Ameca Desktop, Ami and Azi) extends the same expression engine to smaller form factors, with Generation 3 hardware appearing during 2025 and 2026.

Ameca is usually shown as a torso on a stand: standard configurations do not walk, and its arms are for gesturing, not manipulation. With GPT-class language models wired to its face, it delivers the best conversational embodiment currently demonstrated anywhere.

Delivery evidence

Announcement is not arrival.

Commercially deployed

Deployed at venues, expos and institutions worldwide since 2022; documented US and international customers

Reality check

Ameca is what it appears to be, which is rare in this database: a superb expression platform with real customers. The honest caveats: conversation quality depends on the LLM configured for it; it cannot manipulate objects usefully; and event appearances sometimes mix autonomous conversation with operator-triggered responses, which venues do not always disclose.

Capability ledger

What has actually been shown?

3 material claims are separated by evidence setting and human involvement. The strongest recorded signal is Commercially deployed; it does not automatically transfer to every row.

01
conversation

Open-ended spoken conversation with lifelike facial expression

Countless unscripted press and public interactions since 2022

Independently testedVerified July 2026
Control mode
Autonomous
Human intervention
None in LLM mode; venues can also run scripted or operator-driven sessions
02
entertainment

Star attraction at museums, expos and events

Commercially deployedVerified July 2026
Control mode
Mixed / assisted
Human intervention
Venue-dependent mix of autonomous and operator modes
03
education

Classroom and university engagement programs

Real-world demoVerified June 2026
Control mode
Mixed / assisted
Human intervention
Facilitated sessions
Body & movement

Hardware record.

Height
187 cm
Weight
49 kg
Locomotion
Stationary torso on stand (standard); research walking versions not sold
Payload
Not applicable; arms for gesture only
Runtime
Mains powered; continuous operation
Charging
Mains powered
Top speed
Not applicable
Degrees of freedom
Head/face 32 DoF class; 51+ total in full configuration (company figures)
Hands
Articulated fingers for gesture, not manipulation
Face / expression
Best-in-class: micro-expressions, eye mechanisms with cameras, silicone skin over 17+ facial actuators
Sensors
Eye cameras, chest camera, microphone array, position/force sensing
Intelligence & control

Software record.

AI platform
Engineered Arts Tritium OS; LLM integration configurable (GPT-class models demonstrated)
Voice
Fluent open-ended conversation with lip sync
Languages
Multiple, depending on configured speech stack
Visual recognition
Face detection and tracking, person re-identification, object naming demos
Memory
Session context; persistent memory configurable by integrators
Processing
Local control with cloud LLM calls
Autonomous abilities
Autonomous conversation and expression; no physical task autonomy
Teleoperation
Operator control and scripted performance modes available to venues
Software updates
Tritium platform updates for customers
Developer access
Yes: API and developer tools for owners
Privacy & safety

Who can see, hear, or take control?

Recording storage
Operator/venue controlled; cameras and mics record for interaction processing
Remote operation possible
Yes
Teleoperation disclosure
At the venue's discretion; not standardized
Camera / microphone controls
Yes, under operator control
Data deletion
Venue/integrator responsibility
Account required
Not applicable (B2B)
Emergency stop
Hardware e-stop standard
Children and pets
Stationary; supervised public settings
Security updates
Vendor-supported platform
Open questions

Disputed or unverified.

  1. All published purchase prices are third-party estimates; Engineered Arts quotes per project
Lifecycle record

Status history.

  1. Public demonstration

    Viral reveal video

  2. Commercially available

    CES debut; sales and rentals to venues begin

  3. Commercially available

    US restructuring and $10M Series A to scale production

  4. Commercially available

    Generation 3 desktop line surfaces; education deployments continue