faster-whisper
Whisper transcription reimplemented on CTranslate2 for speed
What it is
A Python reimplementation of Whisper on the CTranslate2 inference engine. Its README reports up to four times the speed of the reference implementation at the same accuracy with lower memory use, and supports 8-bit quantisation on CPU and GPU. For developers adding transcription to Python pipelines.
Good for
- Transcribe long recordings in Python
- Cut GPU memory use for Whisper models
- Run quantised speech recognition on CPU
Install or run
pip install faster-whisperCommand as given in the project README. Read the repository before you run anything on your machine.
What you need
Python; NVIDIA cuBLAS and cuDNN libraries for GPU use
Sources
- https://github.com/SYSTRAN/faster-whisper (project README, read 10 Nov 2026)
- https://api.github.com/repos/SYSTRAN/faster-whisper (GitHub API, read 13 Nov 2026)
The summary is ours, written from the README. Spotted an error? See how we work.