1.3 KiB

name description
signet-speak Speak audibly to the user via text-to-speech using OpenAI's gpt-audio-mini model through OpenRouter. Use when the agent wants to talk out loud, greet the user verbally, deliver information audibly, or when spoken audio output would enhance the interaction. Also triggers when the user asks the agent to "say something", "speak", "talk to me", "use your voice", or "read this aloud".

Signet Speak

Speak out loud via TTS. Uses OpenRouter's openai/gpt-audio-mini with Signet daemon's exec_with_secrets for secure API key injection.

Quick Start

~/.agents/skills/signet-speak/scripts/speak.sh "whatever you want to say"

Override voice (default: ash):

VOICE=coral ~/.agents/skills/signet-speak/scripts/speak.sh "hello there"

Voices

alloy, ash (default), ballad, coral, echo, fable, onyx, nova, sage, shimmer, verse, marin, cedar

Prerequisites

  • Signet daemon running (signet daemon start)
  • OPENROUTER_API_KEY in Signet secrets
  • ffplay, jq, curl on PATH

Troubleshooting

  • No audio data: Check daemon (signet status), verify API key
  • 401 error: Re-store key with signet secret put OPENROUTER_API_KEY
  • No sound from speakers: Test with ffplay -f lavfi -i sine=f=440:d=1 -nodisp -autoexit