[BUG] Cowork voice dictation: mic activates briefly on 2nd attempt, then stops (Windows 11)

Open 💬 4 comments Opened Jun 30, 2026 by gunterhetzner-ui

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?

On Windows 11, voice dictation in Claude Cowork (Desktop app — NOT a terminal/CLI; "Other" selected under Terminal/Shell because there is no Desktop option) works on the first use, but breaks on every subsequent attempt. On the second try the microphone activates briefly (icon turns active) and then the recording session stops on its own after a moment, before anything can be dictated. No error UI is shown.

This looks like an audio-session/stream teardown problem: the first dictation session does not release the MediaStream / AudioContext cleanly, so the next acquisition is torn down immediately after it starts.

What Should Happen?

A second (and every subsequent) dictation should start and keep recording, exactly like the first one, until I stop it.

Error Messages/Logs

No error UI is shown. If reproducible, attach the excerpt from logs/claude.ai-web.log at the moment of the second attempt (watch for a "ScriptProcessorNode is deprecated" warning at click time).

Steps to Reproduce

  1. Open the Claude Desktop app on Windows 11 and switch to Cowork.
  2. Click into the input field and start voice dictation. Speak — works fine.
  3. Stop dictation.
  4. Trigger voice dictation a second time.
  5. Observe: the mic icon activates briefly, then recording stops on its own; no text is dictated.

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

1.15962.1.0

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

Cross-checks already done:

  • Win + H (Windows built-in voice typing) works into the same field → mic hardware and OS permissions are healthy.
  • Microphone access for desktop apps is enabled under Settings > Privacy & security > Microphone.
  • Fully quitting (via Task Manager) and relaunching restores the first dictation; only the second one breaks → points to a per-session teardown issue, not setup/permissions.

View original on GitHub ↗

This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗