Maya AI Voice: What Sesame's Voice Means for Voice Agents

This is part of our AI Customer Service Representative guide.

TL;DR

• Maya AI voice (and her sibling Miles) are Sesame's experimental voices that set a new bar for emotionally expressive speech in 2026.
• Sesame is free to demo but is not a production voice-agent platform — there is no SIP, no CRM, no observability layer.
• For real phone work I run CallSphere's six vertical agents on GPT-Realtime-2 with managed TTS in 57+ languages.
• Explore a live demo and compare current plans to find the right fit for your business.

What is the Maya AI voice

The Maya AI voice — and her counterpart Miles — are Sesame's flagship experimental voices, designed to push the bar on emotional expressiveness, micro-pauses, breathing, and the small disfluencies that make synthesized speech feel less synthesized. Search interest for "maya ai voice" exploded in late 2025 after Sesame's public demo went viral, and the questions I still get from operators in 2026 are mostly: how does it work, is it free, and can I use it for my phone agent.

I am Sagar Shankaran, founder of CallSphere. I run six live AI voice agents in production across healthcare, real estate, sales, salon, after-hours, and hotels. I have spent serious time with the Maya demo, with our own production TTS pipeline, and with the alternatives. Here is the honest read.

Sesame Maya AI: what is actually going on under the hood

Sesame's Maya AI voice is built on a research-grade speech model that emphasizes prosody — the rhythm, stress, and intonation of speech — far more aggressively than typical TTS. The model leans into breaths, pauses, and small vocal artifacts that human listeners associate with natural speech. The result, on the demo at least, is striking. People come away from a 5-minute Maya conversation reporting that it felt closer to talking to a person than any voice they had used before.

The catch, and this is important if you are evaluating it for production: Sesame's Maya is a research and demo experience, not a developer-facing API with SLAs, SIP support, observability, or a CRM integration layer. You can talk to her on the website. You cannot wire her into your healthcare call center on Monday morning.

Is Sesame AI free?

Yes — at the time of writing, the Sesame Maya AI demo is free to use on Sesame's site. You go to the page, click to start, and talk to Maya (or Miles) in your browser. There is no signup wall for the demo, no card, no usage cap that I have hit personally.

What is not free is using Maya in a commercial product, because at the moment Sesame does not offer that as a productized API. If you want production-grade emotional TTS for a phone agent in 2026, your real options are the major TTS vendors (ElevenLabs, Azure, OpenAI TTS, PlayHT) on a managed voice-agent platform like CallSphere, not the Sesame demo wired into your own backend.

Maya and Miles AI: when does a voice receptionist need this?

The honest answer is: rarely. An ai voice receptionist for a dental office, a salon, a real estate brokerage, or a hotel does not need Maya-level emotional expressiveness. It needs to be clear, fast, polite, and accurate. First-token latency under 300ms matters more than dramatic prosody. A reliable booking flow matters more than a perfect breath sound.

That said, there are a handful of use cases where emotional TTS pulls real weight: grief and bereavement intake, long-form audiobooks, mental-health screening, and high-touch sales. For those, I would absolutely take the closest production-API equivalent to Maya I could get. For booking your kid's haircut on a Saturday, the current ElevenLabs and OpenAI voices on CallSphere are already past the threshold of caller comfort.

Maya AI voice chat: how it differs from a phone agent

The Maya AI voice chat experience on Sesame's site is browser-based, full-duplex, and tuned for a single user talking to a single conversational model. There is no SIP, no number, no CRM hookup, no booking, no transfer-to-human escalation, no row in any database after the call ends.

A CallSphere voice agent is the opposite shape. The conversation is the start of the work, not the whole work. After every call we write rows to the calls table, the conversations table, the appointments table, and the crm_events table. Webhooks fire to Stripe, Calendly, the EHR, or whatever CRM the tenant uses. The voice is the interface; the structured outcome is the product.

How CallSphere does this in production

CallSphere runs six live verticals on GPT-Realtime-2 (128K context) with a managed TTS layer covering 57+ languages. We expose 14 function tools across the verticals — book_appointment, check_insurance, transfer_to_human, send_followup_sms, create_ticket, lookup_listing, and so on. Voices are curated per language; for US English we expose four female and four male profiles, all under 300ms first-token TTS latency.

For an ai voice receptionist specifically — healthcare front desk, salon booking, hotel concierge — the receptionist agent is one of the most-deployed shapes on the platform. Setup takes 3 to 5 business days, billing starts at $49/mo Act, and the affiliate program pays 22% revenue share if you bring a customer in.

If Sesame ever releases Maya as a production API with SIP support and predictable pricing, I will evaluate it for our managed pool. Until then, the realistic production answer for a voice receptionist in 2026 is a major-vendor TTS voice on a managed platform, not a research demo.

A real example walk-through

A nine-location pediatric dental group asked specifically about "the voice from that Sesame demo" when they came to CallSphere in March. We walked them through the production realities — no SIP, no CRM, no observability on Maya — and recommended an Azure Neural female US voice on our healthcare agent instead.

They went live in four business days on the $99 Orchestrate tier with HIPAA BAA in place. First 30 days: 5,200 calls handled, 880 new appointments booked, 140 escalations to a human, zero complaints about the voice. The lesson I keep relearning: callers care about being understood and helped, not about whether the voice is the bleeding-edge demo from last quarter.

Pricing and how to try it

CallSphere bundles model, TTS, STT, and transport into one per-interaction price:

• Act $49/mo — one agent.
• Orchestrate $99/mo — popular tier, workflows across multiple systems.
• Custom Build $149/mo — bespoke apps, integrations, and unlimited ongoing custom development; priority support.

Explore a live demo and compare current plans to find the right fit for your business.

Explore a live demo and compare current plans to find the right fit for your business.

Frequently asked questions

What is the Maya AI voice? The Maya AI voice is Sesame's experimental conversational voice, designed to push the bar on emotional prosody — breaths, pauses, and natural disfluencies. Maya and her counterpart Miles are demoable on Sesame's site as a browser-based voice chat. As of mid-2026, Maya is not available as a production API for third-party voice agents, so for real phone work you use major-vendor TTS on a platform like CallSphere instead.

What is Sesame Maya AI? Sesame Maya AI is Sesame's flagship research voice, focused on emotional expressiveness and natural-sounding speech rather than on being a productized TTS API. It is a research and consumer-demo experience. Sesame as a company is building toward what they describe as personal companion AI, and Maya is the most visible piece of that. For developers building business voice agents, Sesame is not currently the right tool — managed platforms like CallSphere are.

Are Maya and Miles AI free to use? Yes, the Sesame demo where you can talk to Maya and Miles is free to use in the browser. There is no card, no signup wall, and no per-minute fee. What is not free is using Maya or Miles inside your own product, because Sesame does not currently offer that as a developer-facing API. For commercial use, you need a production TTS vendor or a managed voice-agent platform.

Is Sesame AI free? The Maya AI voice chat demo on Sesame's site is free. Whether Sesame stays free as the company productizes is an open question. If you are building a business voice agent in 2026 and want predictable cost, CallSphere bundles model and TTS into a per-interaction subscription from $49/mo Act, which is more useful for budgeting than a free demo that might change tomorrow.

Can I use the Maya AI voice for an AI voice receptionist? No — not in production. The Maya AI voice is a demo, not a productized API with SIP, observability, or commercial licensing. For an AI voice receptionist in 2026, the right shape is a managed voice-agent platform with curated TTS voices, real CRM integration, and structured call data. CallSphere's healthcare, salon, and hotel concierge agents all run as voice receptionists today, with setup in 3 to 5 business days.

What is the difference between Maya AI voice chat and a CallSphere voice agent? Maya AI voice chat is a free, browser-based conversation with a research model. A CallSphere voice agent is a production phone agent — it answers a real number over SIP, executes function tools to book appointments and check insurance, writes structured data into Postgres, and integrates with your CRM. The Maya experience is the front half of the iceberg; CallSphere is the rest of it.

Will CallSphere add a Maya-style voice? If Sesame releases a production API with SIP support, predictable pricing, and acceptable licensing, we will evaluate it for our managed voice pool. Until then, CallSphere uses the strongest production-grade voices available from Azure, OpenAI, ElevenLabs, and Google, screened against real call corpora and curated per language. Operators do not pick a TTS vendor — they pick a voice from our list.

Related reading

• How to choose a voice for your AI agent
• AI customer service representative guide
• AI voice receptionist for small business
• Female text to speech voices ranked
• Voice agent latency benchmarks under 800ms