AI
Viture's New AI Glasses Skip the Camera to Map Your Personal Memories
Viture unveiled Vonder, smart glasses without cameras that use bone-conduction sensors and AI to build a personalized memory graph of your daily life and habits.
WIRED — AI
Viture has announced Vonder, a fresh take on smart glasses that ditches the embedded cameras entirely. Instead of relying on visual recording, the frames listen—using bone-conduction sensors and microphones to capture the wearer's own voice throughout the day, creating a digital map of memories, ideas, and patterns without the privacy concerns that have shadowed competing products.
A Pivot From Display-Focused Glasses#
Viture built its reputation on wearable display glasses like the Viture One and Beast—devices designed to project screens from connected devices directly into your field of view. The new Vonder glasses represent a departure, trading the chunky form factor and projected displays for a pair of frames that look and feel like regular eyewear. The glasses can be fitted with prescriptions and come in multiple styles with clear, transition, or tinted lenses, priced at $299 with preorders launching September 28.
How the Memory System Works#
The Vonder's core feature is a built-in AI assistant that listens passively via bone-conduction sensors—which pick up the vibrations of your own voice—and records ambient sound when you trigger a meeting mode. As you talk throughout the day about your activities, the assistant builds what Viture calls a "personal memory graph," tracking recurring topics, places, and patterns. The system then presents this information as an interconnected web in a companion app.
Viture CEO David Jiang demonstrated how the glasses had tracked his life over three months, allowing him to query the data—asking where nearby coffee shops matched his preferences or seeking context for his stress levels. The system even works predictively, surfacing topics Jiang had been thinking about but hadn't yet acted on, such as hiring concerns or announcements he was planning to make.
The AI engine itself pulls from multiple large language models, routing queries to OpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, or DeepSeek depending on what's fastest and most cost-effective for the task.
