Voice AI at the Speed of Conversation

A new generation of low-latency speech models and voice agents is finally keeping pace with human conversation — and the language barrier that has stood between East Asia and the rest of the world is starting to fall away in real time. On Wednesday, June 24, we gathered with Kindred Ventures at their San Francisco office to mark that moment, alongside a room full of engineers, researchers, and builders who care about the same thing we do: conversation that flows naturally, no matter what language it starts in.

The hard problem we're solving

Our co-founders Noriyuki Kojima (CEO) and Jungo Kasai (CTO) opened the evening with a look at Kotoba and why we exist. We're building a frontier voice AI company around a real-time speech model and simultaneous interpretation technology for East Asian languages — Japanese, Korean, Chinese, and English. Simultaneous interpretation is one of the most demanding tasks in all of AI: it isn't enough to be accurate, you also have to be fast, preserve the speaker's voice and intent, and handle the messy realities of real human turn-taking — all at once, with latency low enough that the conversation never stalls. Our Koto foundation model is built for exactly that.

Rather than describe it, we showed it. A live speech-to-speech simultaneous interpretation demo let the room hear what real-time, voice-to-voice interpretation actually feels like when the lag disappears.

Backed for the road ahead

We also shared the news of our recent funding round, supported by Kindred Ventures, Salesforce Ventures, and Sony Innovation Fund. This capital lets us keep advancing the Koto model and bring real-time interpretation to many more people, places, and devices.

Conversations from the frontier

Partnership and the opportunity in Asia.

Jungo Kasai sat down with Will Matteson (Speech AI, Salesforce) and Ludovic Copéré (Sony Innovation Fund) to explore what excites them most about voice AI today, where the biggest opportunities for collaboration lie, and what they each hope we'll be saying about voice AI's impact in Asia three years from now.

The infrastructure and research behind voice agents.

Noriyuki Kojima hosted Mike Henry (Founder & CEO of Parasail, the "AI Supercloud" inference and training platform) and Sewon Min (Assistant Professor at UC Berkeley, BAIR / Berkeley NLP). Their conversation moved between two complementary vantage points — the realities of serving billions of tokens a day and building production voice-agent pipelines, and the research frontier of LLM reasoning and where voice AI is headed next.

A look ahead.

Steve Jang (Founder & Managing Partner, Kindred Ventures) closed the evening with our co-founders, digging into the biggest opportunities and technical bottlenecks for voice AI in East Asia, the architectural questions still open in speech systems — edge or datacenter, cascaded or end-to-end — and our ambitions for this year and beyond.

Thank you

The night wrapped with a reception, hands-on time with the Kotoba app at our demo stations, and — fittingly for a company built around East Asian languages — a Korea vs. South Africa World Cup watch party.

Thank you to Kindred Ventures for hosting, to our guests for sharing their time and insight, and to everyone who came out. Real-time, natural conversation across languages is no longer a someday idea. It's happening now — and we're just getting started.