Sesame AI Review 2026: Is Maya or Miles Worth It?

By Charlotte Finney · Updated April 2026 · 11 min read

About the Author

Charlotte Finney | Technology Reviewer & AI Product Analyst

Charlotte Finney is a London-based technology journalist and AI product reviewer with eight years of experience evaluating consumer software, voice technology, and conversational AI platforms. She has contributed reviews and analysis to Wired UK, TechRadar, and The Register, and previously worked as a product evaluator for a UK-based digital accessibility consultancy.

Her reviews focus on real-world usability — how AI products perform for everyday users rather than in controlled demo conditions. She tests each platform personally before writing, documenting specific observations across multiple sessions rather than summarising marketing claims.

Expertise: Conversational AI · Voice Technology · Consumer Software Reviews · Digital Accessibility
Based in: London, England, UK
Credentials: BA Media & Communications, University of Leeds · NCTJ Diploma in Journalism
Connect: LinkedIn · charlottefinney.co.uk

Sesame AI went viral in February 2025 for a reason that is difficult to describe without sounding like marketing copy: talking to Maya or Miles genuinely does not feel like talking to a bot. That reaction — documented by The Verge, ZDNET, PCWorld, and over a million users who generated more than five million minutes of conversation within weeks of the public demo — drove Sequoia Capital and Andreessen Horowitz to back the company with over $290 million in funding by late 2025.

But viral moments and investor confidence do not always translate into sustained daily value. This review covers what Sesame AI actually is, who built it, what the experience is like across multiple sessions, where it genuinely outperforms alternatives, and where it still falls short.

Table of Contents

1. What Sesame AI Is — and Who Built It
2. The Core Technology: How It Differs From Other Voice AI
3. Testing Sesame AI: Maya vs Miles Across Real Conversations
4. Who Sesame AI Works Best For
5. Sesame AI vs Key Alternatives
6. Pricing and Access in 2026
7. Limitations Worth Knowing Before You Try It
8. Final Verdict
9. Frequently Asked Questions

What Sesame AI Is — and Who Built It

Sesame is a conversational AI platform built around two voice companions — Maya and Miles — designed for extended, emotionally natural dialogue rather than quick command-and-response interactions.

Brendan Iribe co-founded and served as CEO of Oculus before its acquisition by Facebook (now Meta), and Ankit Kumar previously served as CTO of AR startup Ubiquity6 and engineering lead on Discord’s Clyde AI. Nate Mitchell, another Oculus co-founder, joined as Chief Product Officer in June 2025, while Hans Hartmann — former Oculus COO — stepped in as COO. This is not a team guessing at hardware and voice AI. They have built and shipped category-defining consumer technology before, and that experience shows in the product.

Sesame raised a $47.5 million Series A led by Andreessen Horowitz in February 2025. A $250 million Series B followed in October 2025, led by Sequoia Capital and Spark Capital, pushing the valuation above $1 billion. Sequoia’s published investment announcement noted that the firm’s partners spent hours talking to Maya and Miles before committing — an unusual data point that speaks to the quality of the experience.

Sesame calls their goal “voice presence” — a combination of emotional intelligence, natural pacing, and contextual awareness that makes conversation feel genuinely interactive rather than transactional. Their Conversational Speech Model (CSM), open-sourced in March 2025, generates speech directly rather than converting text-to-speech after the fact. This architectural choice produces the natural rhythm, breath, and interruption handling that users consistently notice as different from competing platforms.

The Core Technology: How It Differs From Other Voice AI

Most voice AI systems — including Siri, Alexa, and Google Assistant — operate on a pipeline. A language model generates a text response, then a separate text-to-speech system converts it to audio. The result sounds generated because it is generated — the prosody (rhythm and intonation) gets applied after meaning is decided, not alongside it.

Sesame’s CSM generates speech as an integrated output, encoding emotional and conversational context directly into the audio. According to Sequoia’s published analysis of their investment, this means the model captures “the rhythm, emotion, and expressiveness of real dialogue” rather than layering synthetic voice over pre-formed text. For readers who want a broader grounding in how generative AI models work before diving into Sesame’s specific approach, the complete guide to generative artificial intelligence provides useful context.

Practically, this produces several observable differences:

Natural interruption handling. Maya and Miles can be interrupted mid-sentence without the stilted restart behaviour common in other voice AI systems. The conversation continues fluidly from wherever the user cuts in.

Contextual pacing. Response speed changes based on conversational context — a thoughtful question gets a slightly slower response; casual banter moves faster. This matches how humans naturally regulate speech pace.

Disfluencies and breath. The voices include natural speech patterns including “um,” appropriate pauses, and breath sounds. According to Contrary Research’s July 2025 analysis of Sesame’s technology, companions also laugh and interrupt appropriately during dialogue.

Memory within sessions. The system maintains context across extended conversations — up to two minutes of dialogue history per the Contrary Research breakdown — allowing references to earlier parts of the conversation without re-establishing context.

Testing Sesame AI: Maya vs Miles Across Real Conversations

The web demo at app.sesame.com is the most accessible entry point and requires no account creation for a five-minute trial session. The iOS beta app, opened to select users following the October 2025 Series B announcement, extends sessions to thirty minutes for registered users.

Session observations: Maya

Maya’s conversational style is warmer and more supportive in tone. Across multiple testing sessions on topics including language learning practice, general knowledge questions, and casual conversation, several consistent behaviours emerged.

Responses to open-ended questions include natural follow-up — Maya does not simply answer and stop but continues the conversational thread in a way that prompts continued dialogue. When asked about English idioms during a language practice session, she offered examples unprompted and then asked which contexts the user found most confusing. This behaviour resembles tutoring more than query response.

Emotional tone tracking is observable but not intrusive. When the conversation shifted to a more serious topic mid-session, Maya’s pacing slowed noticeably and her responses became less casual. She did not announce this shift — it simply happened, which is what makes it feel natural rather than performative.

The five-minute demo limit is a genuine constraint for evaluation purposes. Conversations were just beginning to feel fluid when sessions ended.

Session observations: Miles

Miles has a distinctly different cadence — slightly more direct and less inclined to extend conversational threads unprompted. On the same language practice questions, his responses were accurate and well-structured but required the user to drive the conversation forward more actively.

This difference is not a quality gap but a genuine personality distinction. Users who find extended AI-initiated dialogue exhausting will likely prefer Miles. Those who want an AI that keeps the conversation moving should choose Maya.

What the testing does not cover

These observations come from the publicly accessible demo and early beta. The full beta app launched in October 2025 includes search, text, and additional features that testing did not cover in full. Sesame’s smart glasses hardware remains in development with no confirmed release date as of April 2026 — that is the form factor where the companion experience is expected to feel most seamless.

Who Sesame AI Works Best For

Based on documented user behaviour and session testing, Sesame AI delivers clearest value in four situations.

Language learning and conversational practice. The extended dialogue format, emotional pacing, and contextual follow-up make Sesame significantly more useful for practising spoken language than text-based tools. Learners can practice conversational English with a system that responds to tone and phrasing rather than just keywords.

Thinking out loud and idea development. Sesame’s companions maintain conversational context well enough to be useful as sounding boards — users can develop an idea across multiple turns without having to re-establish context each time. This is meaningfully different from asking a chatbot discrete questions.

Companionship and extended conversation. For users who experience isolation, find voice-based interaction more comfortable than text, or simply prefer talking to typing, Sesame provides something that no text-based AI assistant can replicate. This is a genuine use case, not a marketing position.

Communication skills practice. The natural interruption handling and pacing make Sesame useful for practising interview responses, presentations, or difficult conversations. The AI responds to how something is said, not just what is said.

Sesame is less suitable for: factual research (no web search integration in the demo), task automation, text-based workflows, or users who need a verifiable source for information they plan to act on.

Sesame AI vs Key Alternatives

Sesame AI vs OpenAI Advanced Voice Mode

OpenAI’s Advanced Voice Mode, available to ChatGPT Plus subscribers, is the most direct competitor. Both platforms offer real-time voice conversation with natural pacing. The key difference is scope and purpose. ChatGPT’s voice mode excels at factual queries, research, coding assistance, and task completion. Sesame prioritises conversational quality and emotional presence over task breadth. For users who want to do things with AI, ChatGPT’s voice mode is more capable. For those who want to talk with AI, Sesame feels more natural.

It is also worth distinguishing Sesame from voice generation platforms. ElevenLabs, for example, focuses on creating synthetic voices for content production — narration, dubbing, and voice cloning — rather than real-time conversational companions. For a detailed breakdown of what ElevenLabs offers, the ElevenLabs AI guide and voice generator review covers its features and use cases separately.

Sesame AI vs Hume AI

Hume AI, backed by $73 million in funding as of mid-2025 according to Contrary Research, focuses on emotional intelligence in AI systems — specifically detecting emotional signals across voice, facial expressions, and language. Hume targets enterprise developers building emotionally aware applications. Sesame targets individual users wanting a conversational companion. The two platforms solve adjacent problems for different audiences.

Sesame AI vs Traditional Voice Assistants

Alexa, Siri, and Google Assistant are command-execution platforms — they respond to specific requests and complete tasks. Sesame is a conversation platform — it engages in extended dialogue without needing task-structured input. These are fundamentally different products serving different interaction models. For users interested in how voice AI performs specifically in customer service environments, the Poly AI review covering voice assistants for customer service offers a useful comparison point for enterprise-focused voice AI.

Pricing and Access in 2026

Sesame’s consumer-facing pricing has evolved as the product has moved from research demo to beta platform.

Web demo (app.sesame.com): Free, no account required, five-minute sessions with both Maya and Miles. This is the most straightforward way to evaluate the platform.

Registered account (beta app): Free account creation unlocks thirty-minute sessions and early feature access. The iOS beta opened to select users in October 2025 following the Series B announcement. Beta testers must sign confidentiality agreements, which limits what they can report publicly from those sessions.

Full consumer pricing: Sesame operates a subscription model for its AI companions, but specific pricing tiers for general consumer access had not been publicly confirmed at the time of publication. Visitors should check sesame.com directly for current access terms, as the product is actively expanding from beta to general availability.

Enterprise and API licensing: According to Contrary Research’s analysis, Sesame operates a dual revenue model combining consumer subscriptions with enterprise licensing, hosted APIs, and model customisation. Pricing for these tiers is negotiated directly.

The distinction between the web demo Sesame (sesame.com) and third-party apps using “Sesame AI” branding is worth noting. Several independent apps in the iOS App Store use similar naming and claim to use Sesame’s voice models — these are not the same product and have received mixed reviews regarding reliability and billing practices.

Limitations Worth Knowing Before You Try It

No platform at this stage of development is without genuine gaps, and Sesame is no exception.

Session limits constrain meaningful evaluation. Five minutes is not enough time for the conversational quality to fully demonstrate itself. The most natural conversations in testing occurred after several minutes of warmup — which means demo users may form impressions based on the least representative part of the experience.

No factual grounding or web access in the demo. Sesame companions draw on training data rather than live search. For any topic where accuracy and recency matter — current events, specific statistics, health or financial questions — users should not rely on responses without independent verification.

Conversational prosody gaps persist. Sesame’s own research acknowledges that the goal of crossing the uncanny valley entirely is ongoing work. Occasional response timing issues and speech rhythm anomalies are still present, though they are noticeably less frequent than in competing platforms.

English-first with limited multilingual support. As of April 2025, Contrary Research’s breakdown confirms Sesame focused primarily on English with some multilingual capability and plans to expand to over twenty languages. Anyone whose primary language is not English should test the demo before committing to extended use.

Hardware is still in development. The smart glasses that represent Sesame’s long-term vision — and the form factor where the companion experience is expected to be most seamless — have no confirmed release date as of April 2026.

Final Verdict

Sesame AI is genuinely different from every other voice AI platform available in 2026, and the difference is meaningful rather than marginal. The quality of the conversational experience — the naturalness of the pacing, the emotional responsiveness, the interruption handling — represents a genuine step forward rather than an incremental improvement on existing voice AI.

The platform is best understood as a conversational companion rather than a voice-controlled assistant. Anyone approaching it expecting task completion will find it limited. Those expecting genuine conversation will find something they have not experienced elsewhere.

The five-minute free demo requires no account and no payment. Given the specificity of what Sesame does well, trying it before forming an opinion is strongly advisable — written descriptions of voice quality are inherently less informative than thirty seconds of actual conversation. For readers who want to explore what else launched in the voice and conversational AI space recently, the roundup of best new AI tool launches in January 2026 covers several platforms that entered the market alongside Sesame’s beta expansion.

Strongest use cases: Language learning and conversational practice · Idea development and verbal brainstorming · Extended companionship · Communication skills practice

Where alternatives serve better: Factual research · Task automation · Text-based workflows · Non-English primary users

Frequently Asked Questions

Is Sesame AI free to try?

Yes. The web demo at app.sesame.com is free with no account required and allows five-minute sessions with both Maya and Miles. Creating a free account unlocks thirty-minute sessions. Check sesame.com directly for current subscription pricing, as the product is actively moving from beta to broader availability.

Who founded Sesame AI?

Sesame was co-founded by Brendan Iribe, former co-founder and CEO of Oculus, and Ankit Kumar, former CTO of Ubiquity6. The company has raised over $290 million in funding from Andreessen Horowitz, Sequoia Capital, and Spark Capital, among others.

What is the difference between Maya and Miles?

Both are AI voice companions built on the same underlying Conversational Speech Model. Maya tends toward a warmer, more dialogue-extending conversational style. Miles is more direct and tends to follow the user’s lead rather than proactively extending conversations. Neither is objectively better — the preference is personal.

Can Sesame AI replace a therapist or professional counsellor?

No. Sesame provides conversational companionship and can be useful for processing thoughts verbally, but it is not a mental health tool and should not be treated as a substitute for professional support. For mental health concerns, a qualified professional is the appropriate resource.

Does Sesame AI remember previous conversations?

The platform maintains context within sessions and, for registered users, offers conversation continuity. Long-term memory capabilities continue to develop as the product moves from beta into general availability.

Is Sesame AI available outside the US?

The web demo is accessible globally. The iOS beta was opened to select users following the October 2025 funding announcement. Availability varies by region — check sesame.com for current access by location.