openai / openai/codex

[Windows][Voice] Voice disabled after update to 26.901.1978.0

Open
#42,566 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

app bug windows-os
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

Summary

Voice chat stopped starting immediately after the Windows desktop app updated from 26.831.1445.0 to 26.901.1978.0. The current build rejects voice during its readiness check with:

Voice chat is not ready (voice_disabled)

The rejection occurs before the app invokes realtime transport/WebRTC startup.

Environment

  • OS: Windows 11 Home, 64-bit
  • Windows version/build: 10.0.26200 / 26200
  • Current Codex desktop app: 26.901.1978.0
  • Bundled/local Codex runtime: codex-cli 0.153.0-alpha.5
  • Plan: ChatGPT Business
  • Time zone: America/New_York

Steps to reproduce

  1. Launch Codex desktop app 26.901.1978.0.
  2. Open the new voice chat overlay / start a new empty voice task.
  3. Attempt to begin voice chat.
  4. The overlay does not start a realtime session.

This reproduced repeatedly after several full app relaunches.

Expected behavior

A new voice task starts its realtime voice transport and establishes a WebRTC session.

Actual behavior

The desktop log repeatedly records:

[avatar-overlay-realtime] Failed to start realtime in the avatar overlay
errorMessage="Voice chat is not ready (voice_disabled)"

Observed failures: 2026-09-03 11:23-11:25 AM ET.

No thread/realtime/start call follows these failed attempts, indicating the failure occurs in readiness/capability gating rather than during transport setup.

Before/after evidence

Voice worked on the same machine immediately before the update:

  • 10:59:53 AM ET: realtime voice transport and WebRTC sideband started on 26.831.1445.0
  • 10:59:55 AM ET: realtime session ID established
  • Additional successful realtime sessions started at 11:02:16 and 11:03:04 AM ET
  • 11:04:56 AM ET: Windows completed update 26.831.1445.0 -> 26.901.1978.0
  • 11:23-11:25 AM ET: repeated voice_disabled failures on 26.901.1978.0

Checks completed

  • Windows microphone permission for OpenAI Codex: Allow
  • The microphone was successfully used by the prior working voice session at 11:03 AM
  • Two microphone endpoints are present and healthy
  • No other Codex voice task is active
  • Codex account usage is not exhausted
  • Multiple app relaunches did not clear the condition
  • codex features list on the new runtime reports realtime_conversation as under development and false; this may be relevant to a capability-gating mismatch, but I did not change the flag

Assessment

The update is the strongest causal boundary. This appears to be either:

  1. a regression in desktop build 26.901.1978.0; or
  2. an update/server/workspace capability mismatch that the new build evaluates as voice_disabled.

No experimental flags, account state, privacy settings, or installation state were changed while diagnosing.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the avatar-overlay-realtime readiness and capability-gating path, then compare the behavior of builds 26.831.1445.0 and 26.901.1978.0 using codex features list. Verify whether realtime_conversation being false causes the voice_disabled rejection before thread/realtime/start; done means a new voice task reaches realtime transport and establishes WebRTC without that rejection.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
audio-video-rtc, desktop
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.