Sesame AI

Sesame AI is a research lab building conversational voice models. Its Maya and Miles voices run in a free web demo, and the company sells the underlying speech stack to developers.

What is Sesame AI?

Sesame AI is a US startup founded by former Oculus co-founder Brendan Iribe and Ankit Kumar. It ships two speech models, Maya and Miles, that focus on one thing most assistants skip: sounding like a person who is actually listening. The public demo at sesame.com lets you talk to either voice for free in a browser, no account required.

Product Features

• Two voice personas: Maya (female) and Miles (male), each with separate timbre and pacing.
• Low latency: Sesame targets sub-second response so interruptions feel natural rather than clipped.
• Context awareness: the model tracks turn-taking and adjusts timing instead of reading a script.
• Browser demo: runs on desktop Chrome, Safari and Firefox; no plugin or install.
• Developer access: the CSM (Conversational Speech Model) stack is offered for commercial licensing.

Product Highlights

What stands out is the prosody. Maya and Miles pause where a human would pause, and they can be interrupted mid-sentence without the robotic restart you hear from older TTS engines.

The catch: this is not a voice-cloning tool. You cannot clone your own voice, and you cannot export audio files. It is a conversational engine, not a narration generator.

Use Cases

• Testing how a natural voice assistant feels before buying a hosted TTS plan.
• Benchmarking latency against ElevenLabs or OpenAI Realtime voice.
• Prototyping a phone agent where turn-taking matters more than voice cloning.
• Accessibility demos that need a calm, conversational reading voice.
• Research reference for anyone studying speech-to-speech model design.

Is Sesame AI free?

Yes. The web demo is free and unlimited for personal use. Commercial use of the underlying models requires a licensing conversation with the company, and there is no published per-character price.

Quick Summary

• Sesame AI is a voice research lab, not a full TTS product suite.
• Maya and Miles are free to chat with at sesame.com.
• No voice cloning and no audio export.
• Best for conversational agents, weakest for pre-recorded narration.