Announce Mode

Fire-and-forget text-to-speech. The agent speaks a short message aloud — no response expected.

When to Use

• Task completion alerts: "Deploy finished successfully"
• Status updates: "Running tests now, 12 of 47 complete"
• Narration during long operations: "Moving on to the database migration"

MCP Tool

Name: qa_voice_announce Alias: qa_voice_say

Parameters

Parameter	Type	Required	Default	Description
message	string	Yes	—	Text to speak aloud (non-empty after trimming)
rate	string	No	+10%	Speech rate override (e.g., -5%, +15%)

Returns

{
  "content": [{ "type": "text", "text": "[announce] Spoke: \"your message\"" }]
}

Errors

Error	Cause	Fix
Missing message	Empty or missing message param	Provide non-empty message
edge-tts failed	edge-tts not installed	pip3 install edge-tts
Audio playback failed	No audio player found	macOS: built-in. Linux: install mpv

Behavior

1. Text is synthesized via edge-tts to a temporary MP3
2. MP3 plays through system speakers
3. Temp file is cleaned up
4. Returns immediately after playback completes

No mic recording, no session booking, no blocking.

Speech Rate

Default: +10% (snappy delivery for short updates).

Auto-adjusted for long text:

• 300-599 chars: +5%
• 600-999 chars: +0%
• 1000+ chars: -5%

Override with the rate parameter:

qa_voice_announce({ message: "Important update", rate: "-10%" })

Stop Signal

User can end playback early:

touch /tmp/voicelayer-stop

The stop file is cleaned up automatically after detection.