World Premiere: MirrorMe’s Dexterous Hand VIVA Enables Real-Time Human-Robot Four-Hand Piano Duet; VIVA × CADA Milestone Unveiled

Shanghai, China, September 29, 2026 -- MirrorMe today unveiled its dexterous robotic hand VIVA, achieving the world’s first real-time human-robot four-hand piano duet. Moving beyond rote note-for-note playback, the robot can now play with genuine musical expression. MirrorMe also announced the completion of the VIVA × CADA milestone, marking a new frontier in robotic musical intelligence. Powered by this pairing, robots gain capabilities to compose, perform and iteratively learn—evolving from pre-programmed motion executors into interactive performance partners endowed with artistic creation and expressive capacity.

YouTube：https://youtu.be/uE8LBV7190w

The name VIVA derives from the Latin root viv-, meaning “alive” and “vitality.” It also echoes the musical marking vivace, calling to mind lively, spirited renditions, while paying homage to the iconic track Viva La Vida. CADA, the “cognitive core” empowering VIVA, takes its name from the classical-music term cadenza: the virtuosic passage where performers freely unleash musical emotion. Together, VIVA × CADA embodies the Portuguese phrase Viva cada dia – “Cherish every day.”

VIVA Solves the Dexterity “Impossible Triangle”: Hardware Innovation Addresses Long-Standing Industry Bottlenecks

Within dexterous hand development, power, speed and dexterity have long formed a socalled impossible triangle: strength often comes at the cost of speed, while greater degrees of freedom cap the output of individual actuators. MirrorMe set the human hand as its design benchmark and tackled this challenge via forward engineering. The team captured and analyzed performance data from human pianists, distilled core design specifications, then carried out theoretical modelling and topological optimisation of the mechanical architecture. The resulting motion performance approaches the physiological limits of the human hand, supporting intricate hand-shape transitions, multi-finger coordination and fine trajectory control.

The breakthrough uniting all three attributes lies in the world’s first tendon-linkage hybrid transmission system. Tendon transmission relocates actuators proximally and cuts distal mass to deliver high finger speed; rigid mechanical linkages preserve structural stiffness and accurate force transfer to sustain strong fingertip output. This hybrid mechanism faithfully replicates the natural synergies of human hand joints. Every keystroke and finger lift follows motion trajectories mimicking human performers, eliminating stiff, mechanical-looking movement and yielding fluid, natural motion.

Built upon this architecture, the high-performance drive system delivers joint speeds exceeding 1,000° per second, with peak fingertip force reaching 33 newtons. The hand supports rapid fingering shifts as well as powerful fortissimo strikes across the full dynamic range. Challenging techniques including fast scales, trills and repeated notes can be rendered reliably, delivering ample performance margin for tightly synchronised real-time human-robot ensemble playing. The collaborative rendition of Can-Can – a piece making extreme simultaneous demands on robotic strength, speed and fine dexterity – serves as a demonstration of how MirrorMe has cracked this classic impossible triangle.

VIVA × CADA: VIVA as the Dexterous Performer, CADA as the Music Embodied Foundation Model

CADA, MirrorMe’s self-developed music embodied foundation model, gives VIVA its cognitive “mind.” If VIVA constitutes the physical body that brings music to life, CADA is the brain that grants it musical intelligence. The integrated pipeline closes the full loop spanning music comprehension, composition, embodied performance and feedback-driven learning. A general foundation model first interprets natural-language prompts and musical intent. A music generation module then crafts melodies and arrangements. A performance strategy model translates abstract artistic expression into concrete fingering, touch, force and timing commands, which the VIVA dexterous hand executes on a physical piano.

Crucially, each performance marks not an endpoint, but the start of a learning cycle. The system continuously captures acoustic signals, kinematics and real-world playback results, compares outputs against target references, diagnoses deviations and adjusts performance strategies. This establishes a continuous “compose–perform–perceive–evaluate–learn” loop. Rather than executing a static task once, the robot accumulates embodied experience and polishes its technique through physical interaction, rendering the same composition with increasing refinement over time. Viva la Vida: Nueva Vida, premiered at the launch event, was co-created by the foundation model in dialogue with human collaborators, then independently practised and fine-tuned by the robot to shape its own artistic interpretation.

Animating Machinery: From Mechanical Motion to Musical Expression

According to the team, VIVA represents far more than a piano-playing robotic hand. Hardware is inherently cold. Yet once the machine can strike keys, modulate touch force, follow melodic lines and animate music, it unlocks rich expressive possibilities.

MirrorMe views music as a high-barrier scenario weaving together cognition, creativity, expression and precise manipulation—an ideal testbed for exploring deep fusion between foundation models and robotic hardware. General foundation models deliver understanding, reasoning and interactive capability; domain-specific models rapidly translate general intelligence into professional expertise; robotic actuators execute actions in the physical world; ongoing sensory feedback enables continuous skill improvement. This framework charts a key direction for the future of robotics. By starting with music comprehension and live performance, the team aims to equip robots with broader capabilities: to understand, create, act, interact and keep learning.

To grant machines a body; to let cognition emerge from embodiment; and to make that cognition audible. The VIVA × CADA milestone propels robots beyond pre-scripted movement toward artistic expression, transporting artificial intelligence out of digital spaces and into our physical reality.

About MirrorMe

MirrorMe Technology originates from a robotics team at Zhejiang University, building on a decade of technological accumulation. The company focuses on collaborative innovation spanning original “Robot Body – Embodied Model – Ecosystem”.

It has independently developed the world’s only software for full-body coupled dynamics design for robots, substantially expanding performance limits. Its full-size humanoid robot Bolt features golden human body proportions and ushers in an era where humanoid robots reach speeds of 10 m/s.

MirrorMe has proposed the Distributed Embodied Intelligence Architecture (DEIA). This framework connects large language models, vertical foundational models and robot bodies to integrate cognition and motion. Its VIVA × CADA system (dexterous hand × music embodied large model) enables real-time four-hand piano duets between humans and robots for the first time. DEIA endows models with a physical body, allowing ideas to be heard, bringing the beauty of technology into daily life.

Led by Dr. Jin Yongbin, a post-95s researcher, MirrorMe has been featured by major media outlets including CCTV-1, CCTV International, South Korea’s KBS and Brazil’s TV Globo. Having secured multiple rounds of financing, the company’s products are now in mass production and delivery. MirrorMe is undergoing rapid business expansion.

Contact Info:
Name: Ms. Guo Hui
Email: Send Email
Organization: MirrorMe Technology (Shanghai) Co., Ltd
Website: https://www.mirrormetech.com/

Release ID: 89204633

In case of identifying any problems, concerns, or inaccuracies in the content shared in this press release, or if a press release needs to be taken down, we urge you to notify us immediately by contacting [email protected] (it is important to note that this email is the authorized channel for such matters, sending multiple emails to multiple addresses does not necessarily help expedite your request). Our dedicated team will be readily accessible to address your concerns and take swift action within 8 hours to rectify any issues identified or assist with the removal process. We are committed to delivering high-quality content and ensuring accuracy for our valued readers.