What is Sesame
Sesame is a voice AI companion from a San Francisco lab co-founded by Brendan Iribe, the Oculus co-founder, for people who want natural phone-style conversations with a synthetic voice. Its Maya and Miles demos run on the CSM (Conversational Speech Model) with 1B, 3B, and 8B parameter sizes. The startup sits alongside rivals like ElevenLabs in the speech race and frames the work as a step toward assistants that feel less like machines.
Core Features
- Full-duplex dialogue with about 116ms latency, so interruptions feel human.
- Open weights released under Apache-2.0 in March 2025 let you self-host the model.
- Two ready personalities, Maya and Miles, ship with adjustable tone.
- Emotion and pause modeling avoids the robotic monotone of older text-to-speech.
- A free tier via the web demo runs five minutes without an account and 30 minutes when signed in.
- Speaker-adaptive synthesis learns a user's rhythm across a session for steadier replies.
Use Cases
- Language learners practicing casual conversation at their own pace.
- Developers prototyping voice agents on open weights.
- Accessibility users who prefer talking over typing.
- Call-center teams mocking up voice IVR flows before engineering.
Pricing
Sesame has raised $297.5 million from a16z and Sequoia but has not launched paid tiers yet. The web demo is free, there is no subscription fee today, and consumer glasses are planned for 2027. Until then, the open weights give builders a free path to production use.
Our Take
Best for realistic voice experimentation right now; the trade-off is the lack of a commercial API or SLA. Compare with other tools on aifreetool.site/tool-category/ai-chat-tools/.
