[BUG] Voice mode: transcript never inserted after v2.1.202 — 'Listening… Processing…' then nothing (worked in v2.1.201)

Open 💬 0 comments Opened Jul 7, 2026 by gordonzorik-alt

Description

After auto-updating to v2.1.202, voice mode (/voice, hold-space to record) stops delivering text. Recording and transcription appear to run — the UI shows "Listening…" then "Processing…" — but the transcript is never inserted into the terminal input. No error is shown.

Rolling back to v2.1.201 fixes it immediately with no other changes.

Steps to Reproduce

  1. Claude Code v2.1.202 (npm global install) in macOS Terminal
  2. /voice → "Voice mode enabled (hold). Hold space to record."
  3. Hold space, speak, release
  4. UI shows "Listening…" → "Processing…" → then nothing appears in the input line

Expected Behavior

Transcribed text is inserted into the prompt input (as in v2.1.201 and earlier).

Actual Behavior

"Processing…" completes silently; no text is inserted, no error is displayed.

Environment

  • Claude Code: v2.1.202 (broken) / v2.1.201 (working) — installed via npm (@anthropic-ai/claude-code)
  • macOS: Darwin 25.3.0 (macOS 26), MacBook Air, built-in microphone (default input)
  • Terminal.app; mic permission for Terminal granted (verified in TCC)
  • Timeline: worked the night of 2026-07-06 on 2.1.201; auto-updater installed 2.1.202 on 2026-07-07 09:18 local; voice broken on first use after update; npm i -g @anthropic-ai/claude-code@2.1.201 restores it

Notes

  • Mic hardware confirmed working (input level meters active; default input device = built-in mic)
  • Reproduces consistently on this machine on 2.1.202; not tested on other platforms

View original on GitHub ↗