Humanoids Summit· Oct 19, 2026
Designing Interactive Humanoids | Do Park (Lightberry)
Do Park, head of Lightberry, explains at the Humanoids Summit how the company designs interactive humanoids, using its Lumi robot as an example.
Designing the World’s Most Interactive Humanoid | Do Park, Lightberry
Do Park of Lightberry explains how the company designs Lumi for emotionally engaging human interaction at The Humanoid Hub.
In this talk
- 0:46
Why Lightberry built Lumi
Lightberry draws on films and games to understand what makes robots feel lovable and believable. The company began with autonomy software, then added sensors, but found that interactions felt bolted on when the hardware was not designed for them. Lumi uses Unitree's G1 as its locomotion base, with a platform shaped around interaction.
- 3:14
Three parts of emotional connection
Do Park describes appeal, believability, and relatability as the pillars of a strong interaction. Appeal draws someone in; believability makes the robot seem like a character with a life of its own; relatability gives people a reason to return. Pixar's Carl and Ellie from Up show how shapes and props can suggest personality before a character speaks.
- 7:03
Build a character, not just a voice
Appeal can earn an initial conversation, but motion and personality help keep it going. Lightberry designed Rosie around her motivations, fears, and preferences, then aligned her movement with her character. When asked about a favorite show, Rosie named The Midnight Gospel and gave a personal-sounding explanation, reflecting the background the team had written for her.
- 11:02
Make the robot feel attentive
Lightberry wants robots to show that they notice and care about the person interacting with them. Lumi's eyes also serve as cameras, and the team prioritizes responsiveness, concise speech, and face tracking. Do Park says memory and context let the robot recognize returning visitors and shape later interactions.
- 14:46
Deployment, expression, and safety
In Q&A, Do Park said Lightberry aims for an intentionally robotic look to avoid the uncanny valley, while exploring LED eye effects such as blinking and color changes. Early applications include conference entertainment and greeting visitors in buildings, with customizable characters for different deployments. The team also uses spoken cues about low battery or learning to walk to help people keep a safe distance.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
More talks
- Humanoids Summit· Oct 19, 2026
Advanced Bionic Hands for Humans & Robots | Aadeel Akhtar (PSYONIC)
Aadeel Akhtar of PSYONIC presents at Humanoids Summit on bionic hands that serve prosthetic users and help robots learn manipulation tasks.
- Light Origins· Sep 21, 2026
Introducing Light-O1#humanoid #robots #embodiedai
Light Origins’ presenter explains how Light-O1 uses human video to train robot action models and demonstrates it on a Lightwheel setup and a Unitree G1.
- RoboPapers· Sep 4, 2026
Ep#101: Coupled Local and Global World Models for Efficient First Order RL
Joseph Amigo, a second-year PhD student, and Ruhollah, a fourth-year PhD student at NYU, discuss their paper on robot reinforcement learning with RoboPapers hosts.
- Munro Live· Jul 8, 2026
Unitree G1 Humanoid Robot Teardown
Armin Foshnaughski of Munro & Associates, Schaeffler’s Al Mackie and Paul, and Scott Walter of RoboStrategy discuss the Unitree G1’s design and humanoid robotics during a Munro teardown.
- RoboPapers· Apr 27, 2026
Ep#76: OmniXtreme: Breaking the Generality Barrier in High-Dynamic Humanoid Control
Yunshen Wang and Shaohang Zhu, paper authors from BIGAI and collaborating institutions including Unitree Robotics, discuss OmniXtreme’s approach to general high-dynamic humanoid control with RoboPapers hosts and audience.
