gpt-live
GPT-Live now powers ChatGPT Voice for paid and free users
OpenAI released GPT-Live voice models for full-duplex conversation workflows.
Verify official source →Updated July 2026
S
SesameSesame vs GPT-Live: Which AI Voice Assistant Feels More Human?
Sesame delivers the stronger sense of personality and conversational presence, while GPT-Live is the more capable all-purpose voice assistant. Sesame is better for open-ended, human-like conversation; GPT-Live is better for translation, language learning, web-backed answers, reasoning, and multimodal everyday work.
We may earn a commission from qualifying purchases. Learn more
Live change feed
Recent GPT-Live and Sesame changes
Shared with the homepage and tool pages
Best Fit 2026
Best choice depends on use case
Use the verdict below to match each tool to your workflow.
GPT-Live
Freemium4.6
GPT-Live is OpenAI's full-duplex voice model family for ChatGPT Voice. It can listen and speak with more continuous timing, and GPT-Live-1 can delegate search or deeper reasoning to GPT-5.5. OpenAI launched GPT-Live-1 for eligible paid consumer plans and GPT-Live-1 mini for free users; API access was still described as coming soon at our July 17 review cutoff.
Latest verified update · 2026-07-09
GPT-Live now powers ChatGPT Voice for paid and free users
OpenAI released GPT-Live voice models for full-duplex conversation workflows.
Best For
Live translationLanguage learningSearch and reasoningMultimodal ChatGPT workflowsEveryday voice assistance
Pros
- ✓Full-duplex voice supports more natural overlapping speech and turn-taking
- ✓Can delegate web search and deeper reasoning to GPT-5.5
- ✓Strong fit for live translation, language practice, and information-heavy tasks
- ✓Available across ChatGPT web, iOS, and Android
- ✓Published safety documentation and safeguards for younger users
Cons
- ✗Still feels more like a highly capable assistant than a persistent character
- ✗Agora's controlled test found it slower to yield after an interruption than the previous voice mode
- ✗No video or screen sharing at launch
- ✗API access was announced as coming soon rather than generally available
- ✗Launch availability excluded Business, Enterprise, and Edu workspaces
Sesame
4.5
Sesame is a conversational voice preview built around four persistent agents — Maya, Miles, Simone, and Charlie — with distinct voices, personalities, points of view, and separate memories. It combines calls and texting with search, deep research, notes, reminders, and post-call summaries, but remains a younger product with stricter call limits and narrower platform availability.
Best For
Open-ended conversationHuman-like voice presencePersistent personalitiesReflective or companion-style chatsUsers who value conversational chemistry
Pros
- ✓Four agents have distinct voices, personalities, perspectives, and memories
- ✓Independent hands-on reviews repeatedly praise its human-like pacing and conversational presence
- ✓Can search in the background without completely breaking conversational flow
- ✓Supports calls, texting, notes, reminders, deep dives, and post-call summaries
- ✓Incognito mode prevents a conversation from creating new memory
Cons
- ✗Logged-in calls are capped at 30 minutes and anonymous calls at 5 minutes, though a new call can start immediately
- ✗No comparable public controlled benchmark against GPT-Live is available
- ✗Android access remains a beta/preview rather than a fully mature release
- ✗Agent personalities can feel overly eager, and interruption timing is not always consistent
- ✗A stable paid pricing structure and general developer API were not publicly documented at review time
Feature-by-Feature Comparison
| Feature | GPT-Live | Sesame |
|---|---|---|
| Core strength | Intelligent, full-duplex voice assistant | Human-like conversational presence |
| Voice naturalness | Much improved, but generally more assistant-like | Independent reviewers consistently describe it as unusually human-like |
| Search and reasoning | Delegates search and deeper work to GPT-5.5 | Background search and optional deep dives |
| Live translation | Continuous translation is a highlighted launch use case | No equivalent flagship workflow documented |
| Personalities | Nine voices, without Sesame-style fixed characters | Four named agents with distinct personalities and points of view |
| Memory | ChatGPT memory and conversation context | Separate memory for each agent plus incognito mode |
| Multimodal experience | Voice with text, images, search, and interactive widgets; no launch-day video or screen share | Voice, text, search cards, notes, and reminders |
| Interruption and noise handling | Vendor test found far fewer false interruptions, but slower yielding after interruption | Natural rhythm, without a comparable public controlled test |
| Call limits | Plan-specific ChatGPT limits | 30 minutes logged in; 5 minutes anonymous; restart immediately |
| Platforms | ChatGPT web, iOS, and Android | Web and iOS; Android beta/preview |
| Developer / workplace access | API coming soon; not in Business, Enterprise, or Edu at launch | No general developer API documented |
Our Verdict
Our combined judgment is simple: Sesame feels more human; GPT-Live is more useful. Neither wins every category, and there is no standardized independent head-to-head benchmark yet. Pick Sesame for conversational presence and GPT-Live for practical voice work.
Choose GPT-Live if…
Choose GPT-Live if the conversation has a job to do. It is the stronger option for real-time translation, language practice, current information, reasoning, and tasks that benefit from the rest of ChatGPT. Its main advantage is capability breadth, not character.
Choose Sesame if…
Choose Sesame if the feeling of the conversation matters most. Its named agents, persistent personalities, voice performance, and memory create a stronger sense of talking with someone rather than operating an assistant. Accept the preview-stage limits and less mature platform coverage.
We may earn a commission from qualifying signups, at no extra cost to you.
Pricing Compared
GPT-Live (official OpenAI sources, verified July 18, 2026): GPT-Live-1 is included with eligible ChatGPT Go, Plus, and Pro plans, while GPT-Live-1 mini powers the free experience. Limits vary by plan. OpenAI had not announced standalone API pricing because API access was still described as coming soon. Sesame (official Getting Started guide and preview pages, verified July 18, 2026): Sesame was offered as a free preview, including an Android beta/preview. Logged-in mobile and web preview calls are capped at 30 minutes per call; anonymous calls are capped at 5 minutes. Users can begin another call immediately. Sesame had not published a stable paid-plan structure at our review cutoff. Because the products disclose limits differently and Sesame remains a preview, we do not name a pricing winner.
Sources & Methodology
This is an editorial synthesis, not a laboratory score. We prioritized first-party product documentation for availability and features, then used independent hands-on reviews for subjective voice quality. Agora's latency results are useful controlled measurements but come from a real-time communications vendor, so we treat them as vendor research rather than neutral certification. Community reports were checked for recurring themes but were not used as primary evidence. No authoritative standardized Sesame-vs-GPT-Live benchmark was publicly available at the review cutoff.
- OpenAI — Introducing GPT-Live ↗
Official launch details, access, capabilities, limitations, and safety links.
- Sesame — Getting Started ↗
Official guide to agents, memory, search, notes, platforms, and call limits.
- Sesame — Mobile Preview ↗
Official current availability page for the iOS preview and Android beta.
- PCWorld — Sesame hands-on review ↗
Independent long-form assessment of naturalness, background search, usefulness, and emotional-design concerns.
- Ars Technica — Sesame 28-minute hands-on ↗
Independent early test and founder comments on pacing, tone, eagerness, and interruption timing.
- Agora — GPT-Live latency and interruption test ↗
Controlled vendor testing of latency consistency, interruptions, background speech, and packet loss.
- Tom's Guide — GPT-Live translation hands-on ↗
Independent hands-on use of continuous live translation and visual response cards.
Frequently Asked Questions
Is Sesame or GPT-Live more human-like?
Sesame. Independent hands-on reviews and recurring user feedback give it the edge for voice presence, pacing, and personality. GPT-Live sounds substantially more natural than earlier ChatGPT voice modes, but its design remains more assistant-oriented.
Which is smarter, Sesame or GPT-Live?
GPT-Live is the stronger choice for broad capability. It can hand search and deeper work to GPT-5.5 and benefits from ChatGPT's wider reasoning and multimodal system. Sesame can search and run deep dives, but its defining advantage is conversation quality rather than maximum task breadth.
Which is better for language learning and live translation?
GPT-Live. OpenAI explicitly highlights language practice and continuous translation, and independent hands-on testing has demonstrated simultaneous live translation. Sesame can hold natural conversations, but it does not publish an equivalent translation-first workflow.
Is Sesame free?
Sesame was offered as a free preview at our July 18, 2026 review cutoff. Calls were limited to 30 minutes when logged in and 5 minutes anonymously, with immediate restarts allowed. A stable paid-plan structure had not been published.
Can developers use the GPT-Live or Sesame API?
Not generally at our review cutoff. OpenAI said the GPT-Live API was coming soon, while Sesame did not document a general public developer API. Recheck the official sources before planning production integrations.
Is there an authoritative direct benchmark between Sesame and GPT-Live?
No standardized independent head-to-head benchmark was publicly available when we reviewed them. Our verdict combines official capability documentation, independent hands-on reporting, one vendor-run controlled GPT-Live test, and clearly labeled editorial judgment.