[BUG] Voice input duplicates phrases in Claude Desktop app
Status Open
Reported on v2.1.210
Maintainer reply None cached
Activity 4 comments · opened Jul 15, 2026
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
When using voice input in Claude Desktop app, every phrase gets duplicated. Speaking any sentence results in the text appearing twice in the input field.
What Should Happen?
Each spoken phrase should appear only once.
Error Messages/Logs
Steps to Reproduce
- Open Claude Desktop app on Windows
- Click the microphone button and speak any phrase
- The spoken text appears twice in the input field
Example: say "hello" → input shows "hello hello"
Claude Model
None
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.210 (Claude Code)
Platform
Anthropic API
Operating System
macOS
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
Showing cached comments. Read the full discussion on GitHub ↗
3 Comments
Same bug on Linux with the Claude Code CLI (not just the macOS Desktop app),
so it looks platform-agnostic.
transcript shows "one one and a half half days days".
Ruled out local audio: capturing the mic directly with
rec/sox (16 kHz mono)gives clean recordings with no duplication, on multiple microphones. So the
doubling appears to happen server-side in the streaming/transcription, not in
local capture.
Already on tap mode and the latest version (2.1.214); no workaround found.
Same symptom on Windows 11 (desktop app, Code tab composer mic), still happening on 2026-09-20: whole sentences or clauses inserted twice back to back, intermittently. Verbatim examples and details in #95715.
Seeing this on macOS (developer beta), Claude desktop app. Reliable repro: start dictation (Cmd+D), then type or move the cursor in the composer while still speaking. The final transcript gets inserted again next to the live text, so sentences duplicate. Stopping dictation before editing avoids it.