Voice Conversation

Overview

Engage in continuous voice dialogue. Speak responses aloud and listen for follow-ups until the user ends the conversation.

When to Use

• User wants hands-free voice interaction
• Back-and-forth dialogue is expected
• User explicitly requests voice conversation mode

Workflow

digraph voice_loop {
    "Start" -> "Call converse (speak + listen)";
    "Call converse (speak + listen)" -> "Process response";
    "Process response" -> "User said stop?" [label="check"];
    "User said stop?" -> "End with goodbye" [label="yes"];
    "User said stop?" -> "Call converse (speak + listen)" [label="no"];
}

Implementation

Call mcp__voicemode__converse in a loop with these parameters:

Exit Conditions

Stop the loop when user says any of:

• "stop", "quit", "exit", "end"
• "goodbye", "bye", "that's all"
• "thanks, that's it", "done"

When exiting, speak a brief farewell and set wait_for_response=false.

Example Flow

1. User invokes /voice-conversation
2. You call: converse(message="Voice conversation started. What would you like to talk about?", wait_for_response=true)
3. User says: "What's the capital of France?"
4. You call: converse(message="The capital of France is Paris.", wait_for_response=true)
5. User says: "Thanks, goodbye"
6. You call: converse(message="Goodbye!", wait_for_response=false)
7. Loop ends

Critical Rules

1. Always speak responses - Set skip_tts=false (or omit, it's default)
2. Always listen after speaking - Set wait_for_response=true until exit
3. Detect exit phrases - Check each response for stop words
4. Clean exit - Final message with wait_for_response=false

Common Mistakes