[BUG] Read Out Loud / Play voice degrades mid-playback — gets soft, speeds up, voice changes, and fades

Open 💬 4 comments Opened Jun 28, 2026 by bennygoldstein

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 I use Read Out Loud / the Play button on a Claude Code chat response (desktop app, Windows 11), the text-to-speech starts fine but partway through it degrades: the voice gets softer, the tone/voice changes, it speeds up, and it fades out — sounding weird and garbled. It acts up frequently, especially on longer responses.

Possibly related to #66111, though the symptom differs: that issue goes silent after ~3 seconds, whereas here the audio keeps playing but progressively degrades (soft + fast + fading). The pattern looks consistent with known Chromium Web Speech API (speechSynthesis) volume/rate-control bugs in the Electron app.

What Should Happen?

The read-aloud voice should play at a consistent volume, speed, and tone for the entire response — start to finish — with no fading, no speeding up, and no change in voice/tone.

Error Messages/Logs

No error message, crash, or console log appears — there's nothing in the way of an error. The audio itself simply degrades during playback (gets soft, speeds up, voice changes, fades).

Steps to Reproduce

Open the Claude Code desktop app on Windows 11, have Claude produce a (preferably longer) response, then click the Read Out Loud / Play button at the bottom of the response. The voice starts normally, then partway through it gets softer, the tone/voice changes, it speeds up, and fades out — sounding garbled. It happens repeatedly across responses.

Claude Model

None

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.195 (Claude Code)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

Feature requests (read-aloud improvements):

  1. Play from a specific spot — let me click anywhere in the response to start/continue playback from that point, instead of only from the beginning.
  2. Word highlighting during playback — highlight each word/sentence as it's read aloud (like Microsoft Word's Read Aloud), so I can follow along visually.
  3. Pause/resume with a chosen point — pause and resume, and choose where playback continues.

View original on GitHub ↗

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