The average knowledge worker spends about a third of their week in meetings. Most of that time you aren't talking — you're waiting for the one moment your input matters, half-listening while a tab of real work sits open behind the call.
VoiceDouble started as a selfish question: what if something could hold my seat, in my voice, and pull me in only when it actually needed me?
What it actually does
VoiceDouble is a small macOS app. It listens to your Zoom, Meet, or Teams call, runs the conversation through a voice agent loaded with your context, and answers in a clone of your own voice through a virtual microphone the meeting app picks up like any headset.
You always win the mic
An AI talking in your voice in a live meeting is only safe if you can override it instantly. So the human always wins: hold one key and your real voice takes the floor mid-sentence, cutting the agent off. Let go and it resumes. No setting to misconfigure, no lag.
Your double answers on its own — but the moment you hold the push-to-talk key, your own mic takes over instantly. Release to hand it back.
- Your cloned voice, never a generic one
- A profile per meeting type, each with its own knowledge base
- Push-to-talk so you stay in control
- Nothing recorded or stored after the call
“You won't miss another meeting — not because you went, but because something with your voice did.”
— The pitch, in one line
Where it's going
Auto-picking the right profile from the meeting you join, richer knowledge bases, tighter latency. This is v0.1 — the foundation. If that sounds useful, grab it and tell me what breaks.