ASR Voice Recorder
Minimal Linux voice recorder for transcription: mic picker, level/clip meter, Silero VAD silence skipping, mono 16 kHz MP3/Opus/M4A to a default folder
About This Project
ASR-Voice-Recorder is a lightweight audio recording tool designed specifically for users who need to capture speech for transcription services. It automates the process of saving mono, 16 kHz audio files—the industry standard for models like Whisper—in MP3, Opus, or M4A formats. The application includes a peak level meter, microphone selection, and a background-capable interface that minimizes to the system tray for long-term recording sessions.
Built with Python and PySide6, the project utilizes Silero VAD to automatically skip periods of silence, ensuring that recorded files contain only relevant speech data. It interfaces with PulseAudio or PipeWire via command-line utilities to capture raw PCM data, which is then processed through FFmpeg. This tool is ideal for researchers, journalists, and developers who require a streamlined, reliable way to generate lightweight audio files for downstream AI transcription pipelines.