500xEngineer.

All postsSHORT3 min read

GPT-Live 1.1: OpenAI Voice API Mute and Standby Pricing

OpenAI released GPT-Live-1.1 and WebRTC standby on Nov 19, 2026. How active billing, muting, idle timeouts and standby costs work.

Muting your microphone leaves a GPT-Live session billing at the active rate. On Nov 19, 2026, OpenAI released GPT-Live-1.1 in the API and added configurable standby for GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. Two settings let a quiet session enter paid standby.

Video not loading? Watch the Short on YouTube

What OpenAI shipped

The OpenAI API changelog lists GPT-Live-1.1 for Live sessions, v1/live/sessions. Existing Live session configurations can use the model ID gpt-live-1.1.

The model page describes a full-duplex voice model: it can listen while it speaks. OpenAI says that, compared with GPT-Live-1, it transcribes speech more accurately and maintains a more consistent accent and voice across languages.

Standby arrived the same day for both GPT-Live-1 and GPT-Live-1.1 WebRTC sessions. Both models have the same active and standby rates.

How GPT-Live bills

The voice latency and cost guide spells out active time:

Active time includes speech, silence before the idle timeout expires, and time spent waiting for backend work while inference is running.

And muting:

Muting microphone input leaves the session active.

That's why a muted GPT-Live session can keep billing $0.05/min. The microphone control leaves inference active.

The standby guide also says:

If you omit the timeout, inference stays active after the first wake until the session ends.

How to set up standby

When creating a WebRTC session, set both session.wake_policy.type to speech and session.idle_timeout_s to a whole number from 5 to 60. This JSON shows the two settings with a 5-second timeout:

{
  "session": {
    "wake_policy": {
      "type": "speech"
    },
    "idle_timeout_s": 5
  }
}

The wake policy waits for speech before starting inference. The timeout returns the session to standby after that many seconds without an input or output transcript. Keep the microphone track active during standby.

Watch session.paused for entry into standby and session.resumed for inference being ready. The connection and session ID stay the same. When someone speaks, the session resumes and restores conversation history.

What it costs

OpenAI's pricing page lists the same rates for GPT-Live-1 and GPT-Live-1.1:

GPT-Live session state Rate
Active $0.05/min
Standby $0.02/min

Voice duration bills per second. Backend model and tool usage is billed separately. The standby rate applies while inference is paused in a WebRTC session configured for standby.

MY MATH: 10 idle minutes at $0.05/min cost $0.50. The same 10 minutes entirely on standby at $0.02/min cost $0.20. This compares time billed at each rate. It excludes active silence before the idle timeout expires.

With the default startup behavior, creating a WebRTC session bills 15 seconds of voice duration while it initializes. That amount is credited against duration charges once the session starts running. Account for session-creation charges when evaluating reconnects.

The catch

  • Connection type: standby is documented for WebRTC sessions. The GPT-Live guide recommends WebRTC for browser, mobile and server applications.
  • Paid standby: paused inference still bills $0.02/min.
  • Wake delay: after standby, the first response to a short utterance may take longer.
  • Timeout window: silence before the idle timeout expires bills active. Waiting for backend work while inference runs also counts as active time.
  • Reconnect costs: With the default startup behavior, new WebRTC sessions have initialization billing, credited against running duration. Include that charge when deciding whether to disconnect.

For longer gaps when an application can disconnect, OpenAI suggests closing the session and starting another when the user returns. Choose the idle timeout by comparing avoided voice duration with session-creation costs and the delay before voice becomes ready again.

The builder angle: my opinion

I'd build a local business a website voice assistant using GPT-Live over WebRTC, with standby configured and a dashboard of active versus standby minutes. The client can see what they're paying for, including the difference between a muted active session and paused inference.

For a team already using GPT-Live, I'd check whether both standby settings are configured and whether their mute control leaves inference running.

This is my opinion, with no income guaranteed.

Daily email

Get each breakdown in your inbox.

One email each evening, right after the Short goes live: the AI tool it covers, the full write-up with sources, and how to get paid for it. Free, and you can unsubscribe anytime.

Read past emails on Substack

Sources

All postsWatch on YouTube