faster-whisper

Package

Whisper speech recognition on CTranslate2, faster and lighter on CPU and GPU

Price: Free, open source
Access: None, runs locally

About

Python library from SYSTRAN that runs OpenAI's Whisper models on the CTranslate2 inference engine. The README reports up to 4 times the speed of openai-whisper at the same accuracy with less memory, plus 8-bit quantization on CPU and GPU. Audio is decoded with PyAV, so no system FFmpeg is needed.

What you can do with it

Transcribe audio files on a CPU with int8 quantization
Batch-transcribe long recordings quickly on an NVIDIA GPU
Get word-level timestamps and skip silence with the Silero VAD filter

Get started

Install Python 3.9 or newer, then pip install faster-whisper
For GPU, install cuBLAS for CUDA 12 and cuDNN 9

Example

from faster_whisper import WhisperModel

model_size = "large-v3"

# Run on GPU with FP16
model = WhisperModel(model_size, device="cuda", compute_type="float16")
# or run on CPU with INT8
# model = WhisperModel(model_size, device="cpu", compute_type="int8")

segments, info = model.transcribe("audio.mp3", beam_size=5)

print("Detected language '%s' with probability %f" % (info.language, info.language_probability))

for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

Details

Hosting: Runs locally
Available in: Worldwide
Official SDKs: Python
MCP server: None