AutoMeeting Notes

A Python web UI for generating speaker-diarized meeting transcriptions from audio recordings. Utilizing the state-of-the-art Whisper speech-to-text model and Pyannote speaker diarization techniques, the library produces precise transcripts that accurately differentiate between speakers in a meeting or conversation. With multi-language support and an optional command-line interface, AutoMeeting Notes simplifies the transcription process.

Key Features

• Automatic transcription of audio files with speaker diarization.
• Web-based user interface + command line interface.
• Configurable output formatting to suit user requirements.

Installation Steps

• Clone the repository using the following command:

  git clone https://github.com/danielalcalde/automeeting-notes
  

• Navigate to the cloned directory and install the required packages:

  cd automeeting_notes
  ./install.sh
  

• Accepting eula from pyannote in hf.co in pyannote/segmentation-3.0 and pyannote/speaker-diarization-3.1 and add your huggingface token with:

  huggingface-cli login

Usage Instructions

Command Line Interface

automeeting_notes path/to/audio_file --output_dir path/to/output_dir

The command will generate a text file containing the transcribed and diarized meeting notes in the specified output directory.

Web-based User Interface

Launch the web application by executing:

automeeting-webui

Acknowledgements

Inspired by https://github.com/yinruiqing/pyannote-whisper and openai/whisper#264