Verified July 24, 2026 · Unofficial GPT-Live resource site
gpt-live is OpenAI's newest ChatGPT Voice model family. This independent hub explains how Live works, which plans and limits apply, why the GPT-Live-1 API is not open yet, and how developers can build voice agents now with GPT-Realtime-2.1.
This site is not an official OpenAI service. Every time-sensitive guide shows its verification date and direct official sources.
What it is
The important shift is not only more natural speech. GPT-Live listens while speaking, handles interruptions, mixes voice with text and images, and can spend more time on harder questions through selectable intelligence levels.
See how GPT-Live-1 and mini map to paid and Free plans, plus the nine voices and current restrictions.
Compare Live with Advanced and Standard before choosing natural interruption, screen sharing, or turn-by-turn speech.
Check rolling 24-hour allowances for Free, Go, Plus, and Pro, including the two-hour cap on one conversation.
API status
OpenAI has launched GPT-Live inside ChatGPT Voice and provides a form to be notified about GPT-Live-1 in the API. Until official access, pricing, model IDs, and documentation ship, describe API availability as pending.
Open the official notification formRead the verified GPT-Live API statusBuild now
Start with gpt-realtime-2.1 over WebRTC. Compare mini when a lower price and faster response matter more than maximum quality.
Use WebSocket for server-owned sessions or SIP for telephony, then add consent, webhook verification, transfers, and escalation.
Keep the model ID, prompt, voice, reasoning effort, tools, safety rules, transcripts, and cost controls configurable.
GPT-Realtime-2.1 pricing, context, and modalities are documented today. GPT-Live-1 API details are not; wait for a real API contract before migrating.
Resource map
The hub separates consumer guidance from developer contracts and records a verification date for every changing fact. Open the source library when you need the underlying OpenAI page.
Launch post, Voice Help, release notes, system card, API notification form, Realtime model pages, and protocol docs.
Compare GPT-Live with GPT-Realtime-2.1 prices, context, modalities, WebRTC, WebSocket, and SIP.
Use the complete task index for setup, limits, Voice mode choices, troubleshooting, models, and APIs.
Choose a task
Each guide covers a distinct GPT-Live search intent with practical steps, a visible verification date, and official citations.
Start, prioritize, interrupt, redirect, and check Codex agent tasks by voice on macOS or Windows, with paired iOS Remote access.
Open guideStart Live on web, iOS, or Android; pick a voice and intelligence level; use images, background mode, and CarPlay.
Open guideCompare rolling Free, Go, Plus, and Pro allowances and the two-hour limit for one Live conversation.
Open guideChoose Live, Advanced, or Standard based on interruptions, video, screen sharing, custom GPTs, and workflow.
Open guideTroubleshoot missing Live, the wrong mode, microphone permission, background noise, early interruptions, and transcripts.
Open guideUnlike GPT-Live in ChatGPT, the developer API lets you choose a model, price tier, context window, and transport.
Open guide