Oculus co-founder’s Sesame launches hyper-realistic voice AI

Most people treat AI voice assistants like annoying roommates: you yell at Alexa to shut up, roll your eyes at Siri's misunderstandings, and default back to texting because voice just feels too awkward.

Sesame, a new startup emerging from stealth on February 27, wants to fix that. Founded by Oculus co-founder and former CEO Brendan Iribe, Ubiquity6 co-founder Ankit Kumar, and former Meta Reality Labs research engineering director Ryan Brown, the company just unveiled an expressive voice model designed for all-day wearable companions.

Natural banter and instant interruptions

The company launched a web demo featuring "Maya," one of two lifelike voices that seem to actually cross the conversational uncanny valley.

In testing by The Verge, the standout capability wasn't just accurate transcription, but conversational agility. When interrupted mid-story and asked to write "herself" into an improvised fantasy quest, Maya instantly pivoted without hitching, crafting a persona as a trap-building gnome engineer defending a castle. With natural breathing pauses and realistic pacing, the flow felt far more responsive than standard voice bots, even holding up favorably against ChatGPT's Advanced Voice Mode.

Under the hood, Sesame released a mini white paper detailing its model, trained on roughly one million hours of publicly available audio. The team announced plans to open-source the underlying models and expand beyond English to over 20 languages in the coming months.

The Oculus reunion goes after smart glasses

Sesame isn't stopping at a browser demo. The end goal is dedicated hardware: smart glasses meant to be worn all day, delivering spatial audio and allowing the voice assistant to see and experience the world with you.

While Sesame only showed off a few early prototype photos, its pedigree commands attention. Backed by an undisclosed round from former Oculus investors Andreessen Horowitz, Spark Capital, and Matrix Partners, the company brings together top-tier hardware and spatial computing veterans.

Why it matters

For users, voice interfaces have long been stuck in command-and-control mode. You bark an order, wait two seconds, and get an answer. What Sesame is pitching is an actual companion—something you test the limits of for fun rather than necessity. The web demo is free and live right now, and it's well worth five minutes just to feel how fluid real-time interjection has become.

For the tech industry, Sesame's strategy reveals where wearables are really heading. Instead of wrestling with optical displays and overheating AR visors, Iribe's team is betting on audio-first smart glasses paired with frontier conversational AI—directly challenging Meta's Ray-Ban momentum.

That said, building hardware is notoriously brutal. Squeezing all-day battery life and high-end microphones into a stylish frame is a massive hurdle, and relying on one million hours of web-scraped audio could bring future training and licensing headaches. But by pairing an insanely responsive voice model with an experienced hardware crew, Sesame looks like one of the most credible wearable AI plays we've seen yet.