Claude text to speech: voice mode, and speech for Claude Code
Claude can talk in its apps: voice mode lets you speak to Claude and hear it answer. Claude Code, the terminal assistant, has no voice of its own.
This guide shows what voice mode does, and three ways to give Claude Code speech: an MCP server, a spoken alert when a task ends, and a free offline engine.
Voice mode in the Claude apps
Voice mode is a beta feature of the Claude mobile apps, Claude Desktop and the web, on every plan. Click or tap the sound wave icon in a chat, pick a voice and talk; Claude answers aloud, and the transcript stays in your chat history. The voices are a small preset set, chosen in Settings, Capabilities, Voice, where the spoken language is set too. Voice conversations count toward your plan’s usage limits. It is made for conversation, not for producing audio files.
Claude Code has no voice mode
According to Claude’s help center, Claude Code and Cowork offer dictation but not voice mode: you can speak a prompt, but Claude Code does not read its answers aloud. Speech has to come from a tool you connect.
Speech in Claude Code with an MCP server
The SuperClonica MCP server gives Claude Code a speak tool with about 190 voices, from stock narrators to game and film characters. Create a free key on the speech API page and run: claude mcp add --transport http superclonica https://superclonica.com/mcp --header "Authorization: Bearer sc_YOUR_KEY". Then ask, for example: voice this paragraph in a calm narrator voice. Claude Code finds the voice and returns a link to the audio file. A free key voices 3 minutes a day.
A spoken alert when Claude Code finishes
Claude Code hooks run a command on events such as Stop, when Claude finishes a reply. Make a short clip once with the free engine: superclonica --text "Done!" --voice F2 --lang en --output done.wav, and put it in a sounds folder. Then add a Stop hook to settings.json: {"hooks": {"Stop": [{"hooks": [{"type": "command", "command": "afplay ~/sounds/done.wav"}]}]}}. On Linux use paplay instead of afplay; on Windows use powershell -c (New-Object Media.SoundPlayer 'C:/sounds/done.wav').PlaySync(). Every finished task now says "Done!" in the voice you chose, offline.
Fully offline speech
The free SuperClonica engine runs on an ordinary processor: python -m pip install superclonica-cpu. Claude Code can run it as a terminal command, for example to voice a README or a whole chapter into MP3 with superclonica --text-file chapter.txt --voice M1 --lang en --output chapter.mp3. Nothing leaves your computer and there is no daily limit.