Voice Conversations with Minnie

Date added: Feb 15, 2026
Status: Parking lot / future exploration
Priority: Nice-to-have, not urgent

Concept

Have actual phone conversations with Minnie instead of just text chat.

Why It's Cool

• More natural interaction (talk while driving, walking, etc.)
• Faster than typing for complex discussions
• Can multitask (hands-free)
• More personal/human

What's Needed

Technical Stack

1. Voice calling API - Twilio or similar

   • Inbound/outbound calling
   • ~$0.014/min cost

2. Speech-to-text - Convert your voice → text

   • Whisper API (OpenAI) or local Whisper
   • Real-time transcription

3. Text-to-speech - Convert my responses → voice

   • ElevenLabs (already have)
   • Low latency for conversation flow

4. Conversation orchestration

   • Real-time back-and-forth
   • Handle interruptions
   • Maintain context across conversation

Integration Options

Option A: Twilio + OpenClaw

• Twilio handles call routing
• Webhook triggers OpenClaw session
• Stream audio → transcribe → respond → synthesize → stream back

Option B: Quo + Voice API (if they add it)

• Wait for Quo to expose voice API
• Simpler integration with existing setup
• May never happen

Option C: Custom VoIP

• Build on SIP/WebRTC
• Most complex, most flexible
• Overkill for single-user case

Estimated Effort

• Setup: 4-6 hours (Twilio integration, real-time audio pipeline)
• Refinement: 2-4 hours (latency optimization, interruption handling)
• Total: 1-2 days of focused work

Cost

• Twilio: $0.014/min ($0.84/hour of conversation)
• Transcription: Minimal (Whisper API ~$0.006/min or free if local)
• TTS: Already covered (ElevenLabs)
• Estimated monthly: $20-50 depending on usage

When to Build This

• After Tier 2 graduation (proven reliability + send authority)
• When voice interaction would genuinely save time vs typing
• When you have 1-2 days for me to build it properly

Notes

• Quan mentioned this on Feb 15 when asking about Quo voice calling
• Not urgent, just "would be cool"
• Parking lot for future

---

Status: Logged, not scheduled. Will revisit if/when it becomes a priority.