Summary
In Codex Desktop, when a side chat is open, pressing the voice dictation shortcut starts voice input in both the main chat composer and the side chat composer.
Only the currently focused/active composer should enter dictation mode.
Environment
- App: Codex Desktop for macOS
26.513.31313 (2867)
- Bundle ID:
com.openai.codex
- Platform: macOS
26.3.1 (a) / build 25D771280a
- Architecture: Apple Silicon /
arm64
Reproduction
- Open Codex Desktop.
- Open a normal/main chat.
- Open a side chat so that both the main chat and side chat are visible.
- Press the voice dictation shortcut (
Ctrl+M).
Observed behavior
Both the main chat and the side chat start voice dictation at the same time.
This makes it unclear which composer owns the recording/transcription result, and it can accidentally capture input into the wrong chat context.
Expected behavior
Only one composer should start dictation:
- If focus is inside the side chat composer, only the side chat should start dictation.
- If focus is inside the main chat composer, only the main chat should start dictation.
- If focus is ambiguous, Codex should pick the active composer or ignore the shortcut rather than starting multiple dictation sessions.
Duplicate search performed
I searched existing openai/codex issues for combinations of:
"side chat" voice
"side chat" dictation
"side chat" "voice dictation"
"Ctrl+M" "side chat"
"voice" "side chat"
"voice dictation"
"Ctrl+M"
I found related side-chat issues and related dictation issues, but did not find an issue for this specific behavior where both the main chat and side chat begin dictation from a single shortcut press.
Summary
In Codex Desktop, when a side chat is open, pressing the voice dictation shortcut starts voice input in both the main chat composer and the side chat composer.
Only the currently focused/active composer should enter dictation mode.
Environment
26.513.31313 (2867)com.openai.codex26.3.1 (a)/ build25D771280aarm64Reproduction
Ctrl+M).Observed behavior
Both the main chat and the side chat start voice dictation at the same time.
This makes it unclear which composer owns the recording/transcription result, and it can accidentally capture input into the wrong chat context.
Expected behavior
Only one composer should start dictation:
Duplicate search performed
I searched existing
openai/codexissues for combinations of:"side chat" voice"side chat" dictation"side chat" "voice dictation""Ctrl+M" "side chat""voice" "side chat""voice dictation""Ctrl+M"I found related side-chat issues and related dictation issues, but did not find an issue for this specific behavior where both the main chat and side chat begin dictation from a single shortcut press.