Enable voice I/O in Claude Code sessions: speak explanations, tool calls, updates, and findings aloud using <say> tags via local proxy; hear user voice input; auto-verify authentication on session start or resume.
npx claudepluginhub cleanser-labs/claudioGive Claude a voice with real-time TTS.
https://github.com/user-attachments/assets/f535dcd9-23c2-46df-b77e-3df8ddf65284
uv tool install git+https://github.com/cleanser-labs/claudio
claudio init # picks the best TTS for your system
claudio # launches Claude Code with voice
git clone https://github.com/cleanser-labs/claudio.git
cd claudio
uv sync
claudio # Launch Claude with voice
claudio --speak text # Speak all text, skip code blocks
claudio --tts kokoro --voice nova # Pick backend and voice
claudio --speed 1.2 # Faster speech
claudio --persona narrator # Use a persona
claudio voices # List available voices
claudio try --voice nova # Audition a voice
claudeYou can run the proxy standalone and point any Claude Code session at it:
# Terminal 1: start the proxy
claudio proxy --speak text
# Terminal 2: launch claude with the proxy
ANTHROPIC_BASE_URL=http://127.0.0.1:9000 claude
The proxy intercepts Claude API traffic, extracts text, and speaks it through your configured TTS engine. All other API behavior is unchanged.
| Mode | Flag | Behavior |
|---|---|---|
tags | --speak tags | Only speaks text inside <say>...</say> tags (default). Claude needs a system prompt telling it to use <say> tags — claudio injects this automatically, but with raw claude you'll need to add it yourself. |
text | --speak text | Speaks all text automatically, skips code blocks and JSON. No special prompting needed. |
all | --speak all | Speaks everything including code blocks. |
off | --speak off | TTS disabled, proxy just forwards requests. |
For most setups with raw claude, --speak text is the easiest since it requires no prompt changes.
Share bugs, ideas, or general feedback.
Voice conversations with Claude Code using local speech-to-text and text-to-speech
Persistent memory system for Claude Code - seamlessly preserve context across sessions
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Intelligent prompt optimization using skill-based architecture. Enriches vague prompts with research-based clarifying questions before Claude Code executes them