faster-whisper is an open-source Python library that reimplements OpenAI's Whisper speech recognition model on CTranslate2, an inference engine for Transformer models. It was released in 2023 by the machine translation company SYSTRAN and is distributed under the MIT license.
It uses converted versions of the same Whisper model weights as the original PyTorch implementation, reducing memory use through techniques such as 8-bit quantization and transcribing faster on CPUs and NVIDIA GPUs. It supports word-level timestamps and a voice activity detection (VAD) filter, and is widely used in subtitle generation and speech transcription tools.