whisper-timestamped
Whisper-timestamped provides an advanced multilingual speech recognition solution with precise word timestamps and confidence scores. Utilizing Dynamic Time Warping, it enhances Whisper models for better speech segment estimation. Integrating Voice Activity Detection, it minimizes hallucinations and handles disfluencies robustly without additional neural networks. Compatible with all versions of the OpenAI-whisper Python package, this tool is optimized for memory efficiency, even with long audio files, enhancing transcription accuracy and supporting multilingual use.
