Sesame
Sesame · Voice & Language · Chatbot
Sesame is an AI voice companion app from the Oculus founding team. Talk to agents like Maya and Miles in natural speech, and let them search the web, take notes, and set reminders.
About Sesame
What Is Sesame
Sesame is a voice-first AI assistant built by a team that includes Oculus co-founder Brendan Iribe. Instead of typing prompts, you speak, and the agent answers in a voice that carries tone, pauses, and emotion rather than reading replies in a flat robotic cadence. The company calls this "voice presence" and treats it as the whole point of the product.
The problem it targets is real. Most voice assistants handle commands well and conversation badly. They sound like machines reading text, so people stop using them once the novelty fades. Sesame goes the other way: its Conversational Speech Model, or CSM, listens to the back-and-forth of a conversation and shapes how words land, which is why the voice feels closer to a person than a command line. The company released CSM-1B, its base speech model, under an open license in March 2025.
The trade-offs matter as much as the upside. The core model was trained mostly on English audio, so non-English performance is uneven. The app is also young: it began as a research preview, and features like task handling and app connectors are still arriving. If you want a polished productivity suite today, this is not it yet.
Getting Started
- Open app.sesame.com in a browser to try the free preview, or install the Sesame app on iOS or Android.
- Pick an agent. Maya and Miles come from the early demo, and newer builds add Simone and Charlie, each with a distinct voice and personality.
- Start talking. There is no prompt box to fill, so you speak the way you would to a person.
- Ask for something concrete, like a quick search, a shopping note, or a reminder, and the agent works it into the conversation.
- Sign in for longer sessions if the free preview cuts you off, and adjust the settings as newer tools appear.
Product Information
A quick look at Sesame's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Everyday voice users
- People testing expressive AI
- Developers and researchers
Tasks
- Think-out-loud conversations
- Quick lookups mid-conversation
- Capturing notes and reminders
- Summarizing a discussion
Scenarios
- Practicing a language or a pitch out loud
- Killing dead time in the car or on a walk
- Prototyping a voice feature
Key features
Conversational Speech That Carries Emotion
The core of Sesame is its Conversational Speech Model, a single-stage system that reads the history of a conversation and adjusts tone, timing, and emphasis as it speaks. It was trained on roughly a million hours of mostly English audio, and the company reports that listeners could not reliably tell its output from human speech when no context was given. That gap reopens once a listener hears the surrounding conversation, which is why the team keeps pushing on prosody.
Named Voice Agents With Distinct Personalities
Sesame ships a small cast of agents rather than one neutral voice. The early preview centered on Sesame Maya and Sesame Miles, with Maya warmer and Miles drier, and later builds add Simone and Charlie so you can pick the tone that fits. Each voice behaves like an AI voice agent with its own personality rather than a style preset. Each agent keeps its own voice and point of view, so switching feels like changing who you talk to, not adjusting a slider.
Web Search, Notes, and Reminders Inside the Conversation
Sesame agents do not just reply. Ask a question and the agent can search the web, and newer builds create notes, set reminders, and summarize takeaways as the conversation moves. The point is to keep you talking instead of switching to another app, so the agent handles the small tasks that would otherwise interrupt you.
Background Tasks and App Connectors
Recent builds add an Activity tab where you hand off a job and let the agent work in the background. Tasks and Schedules cover one-off jobs like trip planning and recurring ones like a morning briefing, and installed Apps connectors for Gmail, Google Calendar, and Google Drive let the agent read a calendar or draft from your inbox. This side of the product is still taking shape, so expect the surface to shift.
Custom Skills Triggered by Voice
Sesame is building a Skills system where you write a short routine and trigger it with a phrase. A skill carries a name, a note on when to use it, and the steps to follow, which follows the same open format Claude popularized. In practice you might say a phrase and have the agent check your calendar, scan your inbox, and read out the urgent items without you touching a screen.
An Open Base Model for Developers
Sesame released CSM-1B, its base speech model, under an Apache 2.0 license, so anyone can run it, study it, or build on top of it. The model has a billion parameters and takes text and audio as input. It is a base model, not the tuned version behind the named agents, so results vary, but it gives developers a starting point they can host themselves.
Pros and cons
Pros
- Voice quality stands out: the speech carries tone and pauses, so long conversations feel less like reading and more like talking.
- Trying it costs nothing: the browser preview and the mobile app are free to start, so you can judge the voice before spending anything.
- The speech model is open source: CSM-1B ships under Apache 2.0, which gives developers a real foundation to build on.
- More than one personality to choose from: four named agents with distinct voices let you match the tone to the task.
Cons
- Non-English support is weak: the model was trained mostly on English audio, so conversations in other languages can sound off.
- Task features are unfinished: connectors, scheduled tasks, and skills are still rolling out, so you cannot rely on them as a finished workflow.
- Paid tiers are coming: the app is free now, but the company has signaled paid plans, so budget for a subscription later.
- The self-hosted model is a base version: CSM-1B is not the tuned build behind Maya or Miles, so your own setup will not sound identical.
Frequently asked questions
Sesame is a conversational AI assistant built around voice, made for talking things through. You speak instead of typing, and the agent answers in a natural voice while it can search the web, take notes, and set reminders, or run background tasks in newer builds.
Related content
Explore related tools, skills, and articles for Sesame.
Sesame Alternatives
Foley
Hanamori Labs, LLC · Voice & Language · Productivity · AI AssistantFoley is a voice to-do list that turns a spoken sentence into a filed task, complete with project, date and time, then helps you plan the day around what is already there.
HireKey
HireKey · Voice & Language · Productivity · LeaningHireKey is an AI job search copilot built by a former tech recruiter. It checks role fit, tailors resumes, runs interview practice and tracks every application.
Fluence
Yay Software (Yannick Huchard) · Writing · Voice & Language · ProductivityFluence is an Android app that reads PDFs, EPUBs, Word files and web articles aloud, follows each sentence as it speaks, and answers questions about the text on the phone itself.