Your AI should do more than chat. It should show up to your meetings. How it works: • Google Meet → audio in • Whisper STT → speech-to-text (90+ languages) • DeepSeek → reasoning and replies • XTTS-v2 → answers in your cloned voice • Replies live in the meeting — like a real participant No paid API for the voice. No cloud lock-in. 100% local inference. Built for founders, consultants, and teams who want an AI teammate that sounds like them — not another chatbot.
I built this because I was tired of AI that only answers in a chat box. The voice was the missing piece — so I cloned my own (30 seconds of audio) and gave the agent a seat in my meetings. It listens, waits for natural pauses, and answers in my voice with an Islamic greeting. The whole pipeline runs free on a Colab GPU: Whisper + DeepSeek + XTTS-v2.
Open source, MIT licensed. What's next: Arabic call-center demo mode and a WhatsApp integration so the same voice answers your clients after hours.
Star the repo if you'd trust an AI in your meeting: https://github.com/Amz34/voice-c...