with one click
help
Show help for claude-talk voice chat commands.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Menu
Show help for claude-talk voice chat commands.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Start a voice conversation with Claude. Launches whisper.cpp STT and audio capture. macOS only.
Spawn or manage voice teammates. Each teammate is a separate Claude personality in its own tmux pane.
Install claude-talk and personalize your voice assistant. Sets up dependencies, then lets you choose a name, voice, and personality. macOS (Apple Silicon) only.
Set system volume to a specific level (0-100) during voice chat
Quick single voice exchange. Speaks a prompt, captures one response via TTS. No persistent session needed.
Stop voice chat. Releases the current voice session and stops the audio server if no other sessions are active.
| name | help |
| description | Show help for claude-talk voice chat commands. |
| disable-model-invocation | true |
Display this help text to the user:
claude-talk - Voice conversation with Claude Code (macOS only)
/claude-talk:install # One-time setup + personalization
/claude-talk:start # Start talking!
| Command | Description |
|---|---|
/claude-talk:install | Install dependencies + choose name, voice, and personality |
/claude-talk:start | Start voice chat with continuous listening |
/claude-talk:stop | Stop voice chat session |
/claude-talk:mute | Pause microphone capture |
/claude-talk:unmute | Resume microphone capture |
/claude-talk:chat | Quick single voice exchange |
/claude-talk:config | View/edit settings (e.g., /claude-talk:config VOICE=Karen) |
/claude-talk:personality | Manage personalities (list, create, switch, edit, delete) |
/claude-talk:teammate | Spawn voice teammates |
/claude-talk:help | Show this help |
Settings are in ~/.claude-talk/config.env. Key settings:
AUDIO_DEVICE - Mic index (find with python3 -c "import sounddevice; print(sounddevice.query_devices())")MIC_GAIN - Mic gain multiplier (built-in mic needs ~8.0, USB mics ~1.0)KOKORO_VOICE - Kokoro TTS voice ID (e.g., bm_daniel, af_heart)KOKORO_SPEED - TTS speed multiplier (default 1.0)Manage multiple personalities with /claude-talk:personality:
/claude-talk:personality — list all saved personalities/claude-talk:personality create — create a new personality/claude-talk:personality switch <name> — switch active personality/claude-talk:personality edit [name] — edit a personality/claude-talk:personality delete <name> — delete a personality/claude-talk:personality export [name] — print personality file/claude-talk:personality import — import from markdown or filePersonalities are saved in ~/.claude-talk/personalities/. The active one is copied to ~/.claude-talk/personality.md.
Once teammates are spawned, they can send each other text messages:
claude-talk message send <personality> "Hey, how's it going?"
claude-talk message broadcast "Listen up team!"
Messages arrive in the recipient's session as:
Teammate <name> said: <text> (direct message)Teammate <name> said to the team: <text> (broadcast)Requires BlackHole 2ch virtual audio device:
brew install --cask blackhole-2ch+ → Create Multi-Output DeviceBlackHole is auto-detected. When available, you can interrupt TTS mid-sentence by speaking.
AUDIO_DEVICE index and MIC_GAINMIC_GAIN or adjust VAD_AGGRESSIVENESS (0-3)