faster-whisper
Whisper reimplemented on CTranslate2 — up to 4× faster with less memory.
- Stars
- 25.9K
- Forks
- 2.1K
- Open issues
- 24
- Last commit
- 2026-10-13
Live from the GitHub API · updated 2026-10-15
What is faster-whisper?
faster-whisper runs Whisper models through the CTranslate2 inference engine, cutting time and memory use with the same accuracy. It supports 8-bit quantization on CPU and GPU and built-in voice activity detection.
Key features
- Up to 4× faster than openai/whisper
- INT8 quantization on CPU and GPU
- Silero VAD filtering
- Word-level timestamps
System requirements
- Platforms
- Windows · macOS · Linux
- GPU
- Optional; NVIDIA with cuBLAS and cuDNN for GPU
- Python
- 3.9+
How to install faster-whisper
Summarized from the project’s README. Projects change quickly — check the GitHub page for the latest instructions.
pip
- 1Install.
pip install faster-whisper - 2Transcribe in Python.
from faster_whisper import WhisperModel model = WhisperModel("large-v3", device="cuda", compute_type="float16") segments, info = model.transcribe("audio.mp3", beam_size=5) for s in segments: print(f"[{s.start:.2f}s -> {s.end:.2f}s] {s.text}")
Strengths
- Much faster than the original
- No ffmpeg dependency
Limitations
- Library only — no UI
faster-whisper alternatives compared
| Software | Stars | License | Platforms | VRAM |
|---|---|---|---|---|
| faster-whisper Whisper reimplemented on CTranslate2 — up to 4× faster with less memory. | 25.9K | MIT | — | |
| Whisper OpenAI’s open-source speech recognition and translation model. | 110.5K | MIT | ~1 GB (tiny) to ~10 GB (large) | |
| WhisperX Fast transcription with word-level timestamps and speaker diarization. | 24.6K | BSD-2-Clause | — | |
| whisper.cpp Whisper in plain C/C++ — fast on CPUs, Apple Silicon and mobile. | 54.4K | MIT | — | |
| Coqui TTS Deep-learning text-to-speech toolkit, including the XTTS voice cloning model. | 46.1K | MPL-2.0 | — | |
| GPT-SoVITS Few-shot voice cloning and TTS with a full training WebUI. | 62.8K | MIT | 6 GB+ |
FAQ
Is faster-whisper free?
Yes — faster-whisper is open-source software released under the MIT license. Check the license (and any model weight licenses) before commercial use.
Does faster-whisper run on Windows, Mac and Linux?
faster-whisper supports: Windows, macOS, Linux.
What hardware do I need for faster-whisper?
Optional; NVIDIA with cuBLAS and cuDNN for GPU
How do I install faster-whisper?
1. Install. 2. Transcribe in Python.
What are the best faster-whisper alternatives?
Popular alternatives include Whisper, WhisperX, whisper.cpp, Coqui TTS. See the comparison table on this page.