A Seoul museum just showed that language support doesn't have to wait for a manufacturer.
TL;DR: VIDRAFT, a Korean Pre-AGI AI startup, enabled Korean-language voice control for Boston Dynamics' Spot robot at the Seoul Robot & AI Museum on July 23, 2026 — without altering Spot's hardware or manufacturer firmware. The on-device AI module processes voice commands locally, keeping visitor data off the cloud. Seoul RAIM is already in talks with VIDRAFT to expand the system toward open-ended natural conversation.
On July 23, 2026, VIDRAFT made history at the Seoul Robot & AI Museum (Seoul RAIM) by deploying Korean-language voice control for Spot, the iconic quadruped robot built by Boston Dynamics. The launch marked the first time a third-party AI company independently unlocked native Korean interaction for Spot — no hardware modifications, no firmware changes from the manufacturer required.
The story quickly drew international attention: Canada-based news aggregator Ground News tracked the development across 66 outlets, with 58 U.S.-based sources picking it up and roughly 74 percent of coverage coming from centrist media. That breadth of coverage signals just how significant the robotics and AI communities consider this milestone.
VIDRAFT integrated an on-device AI module directly into Spot's operating environment at Seoul RAIM, giving the robot the ability to recognize and respond to a set of preset Korean voice commands. When a museum visitor says words or phrases corresponding to actions such as "greet," "sit," "praise," "lie down," or "stretch," Spot identifies the instruction and executes the corresponding movement.
A central design principle behind VIDRAFT's implementation is privacy. Because all voice recognition happens on-device — meaning inside the local hardware at the museum — visitor audio data is never transmitted to an external cloud server. The museum itself confirmed that the system was built specifically to keep voice data local and avoid any cloud processing pipeline.
VIDRAFT CEO Minsik Kim framed the achievement as a broader statement about who should control language support for robots. "Robots understanding human language should not be something only the manufacturer can enable," Kim said, underscoring VIDRAFT's thesis that independent customization should be accessible to any institution, in any language, without waiting for an OEM update cycle.
The system was built without touching Spot's underlying hardware or Boston Dynamics' proprietary firmware — a notable technical constraint that VIDRAFT worked within rather than around. Institutions in non-English-speaking markets have historically been at the mercy of manufacturers' localization roadmaps; VIDRAFT's approach demonstrates a path forward that bypasses that dependency entirely.
The deployment at Seoul RAIM is a proof-of-concept with implications far beyond a single museum exhibit. Robotics deployments in education, healthcare, hospitality, and public services across non-English-speaking countries routinely face a localization gap: the hardware arrives, but meaningful human-robot interaction in the local language doesn't follow for months or years — if ever.
VIDRAFT's on-device architecture sidesteps two of the most common objections to language-customization projects: privacy risk and infrastructure complexity. Because the system runs locally, institutions with strict data-governance requirements — hospitals, schools, government facilities — can adopt similar configurations without exposing sensitive audio to third-party cloud providers.
The museum's confirmed plans to consult with VIDRAFT on next-phase development point toward a more ambitious goal: moving from fixed, preset commands toward open-ended conversational interaction. That shift would allow visitors to ask Spot spontaneous questions and receive meaningful responses, transforming the robot from a demonstration piece into a genuine interactive guide. If realized, it would represent a substantial leap in how public institutions deploy robotic AI.
More broadly, the project illustrates how a focused, independent AI company can extend the utility of established robotics platforms in ways their original manufacturers haven't prioritized. For the global robotics market, that's a meaningful signal: language localization doesn't have to be a bottleneck controlled by a single vendor.
Q: How did VIDRAFT add Korean voice control to Spot without modifying the robot?
A: VIDRAFT implemented an on-device AI module that interfaces with Spot's existing operating environment. The solution works within the robot's current hardware and leaves Boston Dynamics' manufacturer firmware completely unchanged.
Q: Is visitor voice data stored or sent to external servers?
A: No. The museum confirmed that the system processes all voice recognition locally on-device, so visitor audio data never leaves the on-site hardware and is not routed through any cloud processing service.
Q: What comes next for Spot's Korean-language capabilities at Seoul RAIM?
A: Seoul RAIM has stated plans to work with VIDRAFT on expanding functionality beyond the current set of preset commands, with the intended next step being open-ended conversational interaction so visitors can ask Spot questions naturally.
Source: Ground News (캐나다) (2026-07-24) — original article