Claude Code Voice Mode Notes

April 30, 2026 · View on GitHub

Watch the video walkthrough on YouTube

View Video View PDF

A running log of observations, friction points, and feature ideas for the Claude Code voice-mode UX, intended as a precursor to formal feature suggestions.

Scope: Snapshot as of 30/04/2026.

Environment:

  • OS: Ubuntu Linux 25.10
  • Desktop: KDE Plasma
  • Display server: Wayland

Why this repo exists

I use voice input continuously, all day, every day, across essentially every app I work in. The rollout of native voice mode in Claude Code was therefore a significant moment for my workflow.

I've previously submitted feature suggestions and have seen the team act on them — I believe I was the person who suggested the tap mode, which was subsequently implemented. Given that track record, I'm doing this audit deliberately: to surface the remaining points of friction so they can be packaged into clear, well-scoped feature requests.

My voice-input baseline

For comparison, outside of Claude Code I primarily use:

  • Voxtype — my own Gemini-based dictation/notetaking app
  • A Deepgram-based dictation app I've built

So the comparisons in these notes are against tools I've built and tuned for myself. Having any native voice input inside Claude Code at all is great — these notes are small frictions noticed in real use, not a complaint about the feature existing.

Repo structure

  • README.md — this file (purpose, scope, environment).
  • notes/ — individual observations, one per file. See notes/README.md for conventions.
  • .claude/skills/capture-note/ — a Claude Code skill for capturing new feedback notes into notes/ while dictating.

Workflow

When dictating a new observation, invoke the capture-note skill (or just say "capture this as a note") and Claude will append a structured entry to notes/.