faster-whisper-gui is a desktop application for transcribing audio and video files using faster-whisper and whisperX models.
The tool addresses the need for an accessible interface to speech-to-text transcription. It wraps the faster-whisper and whisperX libraries, which provide optimized implementations of OpenAI's Whisper model, and exposes their parameters through a graphical interface built with PySide6. Users can transcribe audio or video to multiple output formats including SRT, TXT, SMI, VTT, and LRC. The application includes voice activity detection through Silero VAD, support for the large-v3 Whisper model, and audio enhancement via Demucs for vocal isolation before transcription. It handles batch processing of multiple files and provides model management features including downloading and converting models.
The tool suits developers and end users who need speech transcription without command-line interaction and want fine-grained control over VAD and Whisper model parameters. It is particularly useful for those working with video files or needing output in subtitle formats. The application supports multiple UI languages and themes. Anyone adopting it should be aware that model files must be obtained separately from Hugging Face or other sources, though the software includes utilities to download and convert models.
The project shows active development with implementation of multiple transcription backends and audio processing features. The codebase demonstrates ongoing refinement of the user interface and file handling systems. The tool integrates established external libraries rather than reimplementing core functionality, relying on faster-whisper for the transcription engine and Demucs for audio separation.
