voice_vault

voice_vault enables you to record and archive all your meetings and conversations with ease. Later, search through them with lightning speed using full-text search.

voice_vault is a Ruby program that captures audio from the default input and output sources using FFmpeg and then transcribes it to text using whisper.cpp. Both the audio (mp3) and transcription will be archived in sub-folders with this YYYY-MM-DD_HH-MM-SS schema.

Installation

• Clone the repository
• Configure the program by editing the config.yml file in the ~/.config/voice_vault/ directory.
• Make the script executable: chmod +x voice_vault.rb

Usage

Simply run the script: ./voice_vault.rb

The script will capture audio from the default input and output sources, transcribe it to text, encode it to MP3, and save it to an archive folder. The folder name will be in the format

~/Dropbox/recordings/ find . | head -n 4
.
./2023-04-19_19-00-55
./2023-04-19_19-00-55/transcription.txt
./2023-04-19_19-00-55/recording.mp3

Configuration

The config.yml file contains the following settings:

• whisper_path: the path to the directory containing whisper.cpp
• whisper_model: the name of the model to use for speech recognition
• whisper_threads: the number of threads whisper.cpp will use
• archive_path: the path to the base directory where the MP3 and text files will be saved