faster-whisper
Python library for fast Whisper speech transcription on CPU or GPU
What it is
A reimplementation of OpenAI's Whisper speech recognition model on top of the CTranslate2 inference engine, cutting transcription time and memory use. It supports 8-bit quantization, batched transcription, language detection, segment timestamps, and Distil-Whisper checkpoints, and decodes audio without a system FFmpeg install. Developers building transcription, subtitling, or voice features into Python apps import it to run Whisper models locally.