FYI: Some projects were obsolete and therefore deleted, so, it is possible, a few links will be not found.
Summary
🎙️ P3X Meeting Assistant is a real-time AI speech-to-text application for meetings and conversations — distributed as the pip-installable Python package p3x-meet-assistant — that captures speaker audio, transcribes it live using OpenAI GPT-4o Transcribe (the highest-accuracy speech model available today), and auto-labels each utterance by voice fingerprint with GPU-accelerated speaker diarization (Speaker 1, Speaker 2, ...). It supports 10 European languages out of the box (English, Hungarian, German, French, Spanish, Italian, Portuguese, Dutch, Polish, Czech) and ships a complete browser-based UI with dark/light themes, adjustable font size, and one-click transcript export, with every session automatically saved to a timestamped text file. Captures system audio natively on Linux via PulseAudio/PipeWire, browser-tab audio on macOS and Windows (Google Meet, Zoom, Teams, YouTube — anything with shareable audio), with NVIDIA CUDA GPU diarization on Linux/Windows, Apple Silicon Metal acceleration on macOS, and graceful CPU fallback everywhere. Cross-platform on Linux, macOS, and Windows — install with pip install p3x-meet-assistant, set your OpenAI API key, and open localhost:8088. Released under the MIT license.