faster-whisper SYSTRAN/faster-whisper

Reimplementation of OpenAI Whisper on CTranslate2, up to 4x faster with less memory, with int8 on CPU or GPU.

Creator kit + Compare GitHub ↗ Install guide (official) ↗

faster-whisper at a glance

Checked 15 Oct 2026

faster-whisper runs Whisper models through the CTranslate2 engine; the README claims up to 4x the speed of openai/whisper at the same accuracy. Benchmarks show large-v2 int8 on GPU at ~2.9 GB VRAM and small int8 on CPU at ~1.5 GB RAM. Code is MIT and the converted Systran models on Hugging Face are tagged MIT. pip install; no system ffmpeg needed. Speech recognition only.

Cite this pageRepoQuarry. “faster-whisper.” Checked 15 Oct 2026. https://repoquarry.dev/projects/faster-whisper

How much of ElevenLabs does it replace?

1 of 4 jobs. Local transcription with word timestamps and VAD filter; the README does not mention speaker diarization.

Speech generation Not covered

Voice cloning Not covered

Speech recognition Full replacement

Voice agents Not covered

For speech generation, voice cloning, voice agents, see other ElevenLabs alternatives. You may also need to build an interface, an API server and hosting yourself.

How much of Otter.ai does it replace?

1 of 3 jobs. Transcribes recordings locally; no live meeting capture, speaker labels or summaries.

Meeting transcription Partial replacement

Speaker identification Not covered

Meeting summaries Not covered

For speaker identification, meeting summaries, see other Otter.ai alternatives. You may also need to build an interface, an API server and hosting yourself.

Facts, with sources

Every answer links to where we found it. We read the documentation; we did not install it.

Report a correction

QuestionAnswerSourceChecked

Runs on my machine? Yes, Python library runs on CPU or NVIDIA GPU The efficiency can be further improved with 8-bit quantization on both CPU and GPU. github.com · faster-whisper#faster-whisper ↗ 1d ago

Code licence MIT github.com · LICENSE ↗ 1d ago

Model weights licence MIT (Systran CTranslate2 conversions of Whisper on Hugging Face, e.g. Systran/faster-whisper-large-v3, ungated) When loading a model from its size such as WhisperModel("large-v3"), the corresponding CTranslate2 model is automatically downloaded from the Hugging Face Hub. huggingface.co · faster-whisper-large-v3 ↗ 1d ago

Commercial use Yes, MIT code and MIT-tagged default model weights huggingface.co · faster-whisper-large-v3 ↗ 1d ago

Paid services needed? None required The module can be installed from PyPI github.com · faster-whisper#installation ↗ 1d ago

Hardware CPU works (small model int8: ~1.5 GB RAM in README benchmark); GPU needs cuBLAS and cuDNN 9 for CUDA 12 (large-v2 int8: ~2.9 GB VRAM on an RTX 3070 Ti 8GB). These are benchmark figures, not stated minimums GPU Benchmarks are Executed with CUDA 12.4 on a NVIDIA RTX 3070 Ti 8GB. github.com · faster-whisper#benchmark ↗ 1d ago

Operating systems Windows and Linux GPU library install paths documented; macOS not stated for this project Purfview's whisper-standalone-win provides the required NVIDIA libraries for Windows & Linux in a single archive. github.com · faster-whisper#gpu ↗ 1d ago

Setup effort Easy; pip install faster-whisper, Python 3.9+. GPU use additionally needs NVIDIA cuBLAS/cuDNN Unlike openai-whisper, FFmpeg does not need to be installed on the system. github.com · faster-whisper#requirements ↗ 1d ago

What it costs to run

Software$0

Required paid servicesnone

Optional paid servicesnone listed

Your hardwareCPU works (small model int8: ~1.5 GB RAM in README benchmark); GPU needs cuBLAS and cuDNN 9 for CUDA 12 (large-v2 int8: ~2.9 GB VRAM on an RTX 3070 Ti 8GB). These are benchmark figures, not stated minimums

ElevenLabs: Free tier (10k credits); paid plans from $6/month, or $5/month billed annually (source, checked 15 Oct 2026). Running it yourself costs your hardware, electricity and setup time.

Project activity

26kstars

2.1kforks

52contributors

Last code changetoday

Latest releasev1.2.1 · 11mo ago

ArchivedNo

From GitHub, today. Stars show popularity, not quality.

Using an AI agent?

The same record, as JSON or through MCP.

get_project("faster-whisper")
GET /api/v1/projects/faster-whisper.json

API and MCP docs →