Voice input broken in CLI: shows 'processing' but only inserts spaces, no transcription

Status Fixed / completed

Reported on v2.1.83

Maintainer reply None cached

Activity 7 comments · opened Mar 25, 2026 · closed Mar 26, 2026

Description

Voice input in Claude Code CLI is enabled (voiceEnabled: true in settings.json) but does not produce any transcription. The status bar shows Voice: processing... but only blank spaces are inserted into the input — no actual transcript appears.

Screenshot

Steps to Reproduce

Enable voice input ("voiceEnabled": true in ~/.claude/settings.json)
Launch claude in terminal
Speak into the microphone
Observe the status bar shows Voice: processing...
Only spaces are inserted into the prompt — no transcription

Expected Behavior

Spoken words should be transcribed and inserted into the chat input.

Actual Behavior

Status bar shows Voice: processing...
Only blank spaces are spammed into the input field
No transcript is ever pasted

Environment

Claude Code version: 2.1.83 (reinstalled via curl -fsSL https://claude.ai/install.sh | bash)
OS: macOS (Darwin 24.6.0)
Shell: zsh
sox installed: No (which sox returns not found)
Claude Desktop app: Installed (was previously deleted and reinstalled during debugging)
Microphone: Working (voice works fine in the Claude Desktop app)

Troubleshooting Attempted

Restarted Claude Code CLI
Killed all Claude processes and restarted cleanly
Reinstalled Claude Code CLI (uninstalled npm version, reinstalled via official install script)
Toggled voice off/on with /voice
Reinstalled Claude Desktop app
No voice-related entries found in ~/.claude/debug/ logs

Notes

Voice input works correctly in the Claude Desktop app with the same microphone
The CLI appears to detect audio input (hence "processing...") but fails to transcribe
Debug logs under ~/.claude/debug/ contain no voice/transcription-related entries, making it hard to diagnose
sox is not installed, which may be relevant if the CLI depends on it for audio capture

View original on GitHub ↗

7 Comments

github-actions[bot] · 5 months ago

Found 3 possible duplicate issues:

This issue will be automatically closed as a duplicate in 3 days.

If your issue is a duplicate, please close it and 👍 the existing issue instead
To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

SalzerRefinedLeads · 5 months ago

Please fix this.

YoanWai · 5 months ago

Confirming on WSL2/Linux - same symptoms, different platform.

Environment:

Claude Code: v2.1.83 (also tested 2.1.81)
OS: Ubuntu 24.04.3 LTS on WSL2 (kernel 6.6.87.2-microsoft-standard-WSL2)
Terminal: WezTerm
Microphone: USB (working, verified via PulseAudio)

Behavior:

Holding Space to push-to-talk floods the input with spaces instead of activating recording
When voice does pick up audio (intermittently), the "processing" state triggers but no transcript is injected into the input field
Installing sox with PulseAudio support (sox libsox-fmt-pulse) did not fix the transcription issue

Workaround: Rolling back to v2.1.80 (npm install -g @anthropic-ai/claude-code@2.1.80) restores voice functionality completely. Space PTT works and transcription appears in the input as expected.

This confirms the regression is cross-platform (not macOS-only) and was introduced in v2.1.81.

Related issues: #38730 (WSL2-specific, auto-closing as dup), #35156 (spacebar PTT types spaces)

YeomansIII · 5 months ago

Seeing same here on 2.1.83.

VossProject · 5 months ago

Same issue here - latest CC Version.

andriivaliukh · 5 months ago

It got fixed in 2.1.84. Closing the issue

github-actions[bot] · 4 months ago

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.