代码Apr 2, 2026·2 min read

whisper.cpp — Local Speech-to-Text in Pure C/C++

High-performance port of OpenAI Whisper in C/C++. No Python, no GPU required. Runs on CPU, Apple Silicon, CUDA, and even Raspberry Pi. Real-time transcription.

SC
Script Depot · Community
Quick Use

Use it first, then decide how deep to go

This block should tell both the user and the agent what to copy, install, and apply first.

# Clone and build
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
cmake -B build
cmake --build build --config Release

# Download a model
bash models/download-ggml-model.sh base.en

# Transcribe audio
./build/bin/whisper-cli -m models/ggml-base.en.bin -f samples/jfk.wav

macOS (Homebrew):

brew install whisper-cpp
whisper-cpp -m /path/to/model.bin -f audio.wav

Real-time microphone transcription:

./build/bin/whisper-stream -m models/ggml-base.en.bin
# Speak into your microphone — see text appear in real-time
Intro

whisper.cpp is a high-performance C/C++ port of OpenAI's Whisper speech recognition model by Georgi Gerganov (creator of llama.cpp). It runs entirely locally with zero dependencies — no Python, no PyTorch, no internet connection needed.

The key advantage: it runs efficiently on CPU. Apple Silicon gets 4-8x speedup via Core ML and Metal. NVIDIA GPUs work via CUDA. Even a Raspberry Pi can transcribe audio. Real-time streaming transcription works on modern laptops.

With 37,000+ stars and the same developer behind llama.cpp, whisper.cpp is the go-to solution for local, private speech recognition.

Model Sizes

Model Disk RAM Speed (CPU) Quality
tiny 75 MB ~390 MB ~32x real-time Good for drafts
base 142 MB ~500 MB ~16x real-time Solid accuracy
small 466 MB ~1 GB ~6x real-time Very good
medium 1.5 GB ~2.6 GB ~2x real-time Excellent
large-v3 3.1 GB ~4.8 GB ~1x real-time Best quality

Hardware Acceleration

Platform Backend Speedup
Apple Silicon Core ML + Metal 4-8x
NVIDIA GPU CUDA 5-10x
Intel CPU OpenVINO 2-3x
Any CPU AVX2/NEON Baseline
Raspberry Pi 4 ARM NEON Usable (tiny model)

Language Bindings

whisper.cpp has community bindings for virtually every language:

  • Python: pywhispercpp — pip install pywhispercpp
  • Node.js: whisper-node — npm install whisper-node
  • Go: go-whisper
  • Rust: whisper-rs
  • C#/.NET: Whisper.net
  • Java: whisper-jni
  • Ruby: ruby-whisper

Features

  • Streaming — real-time transcription from microphone
  • Timestamps — word-level and segment-level timing
  • Translation — transcribe and translate to English simultaneously
  • VAD — voice activity detection to skip silence
  • Speaker diarization — basic speaker identification
  • Output formats — TXT, SRT, VTT, JSON, CSV
🙏

Source & Thanks

  • GitHub: ggerganov/whisper.cpp — 37,000+ stars, MIT License
  • By Georgi Gerganov (also creator of llama.cpp)

Discussion

Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.

Related Assets

Pydantic — Data Validation for AI Agent Pipelines

Python's most popular data validation library, essential for AI agent tool definitions. Pydantic enforces type safety in LLM structured outputs, API schemas, and config files.

Script Depot

Open WebUI — Self-Hosted ChatGPT Alternative

Feature-rich open-source web UI for running local and remote LLMs. Open WebUI supports Ollama, OpenAI, Claude API with RAG, tools, multi-user, and mobile-friendly interface.

Script Depot

Docusaurus — Build AI Tool Documentation Sites

Meta's documentation framework used by major AI projects. Docusaurus creates fast, SEO-friendly docs sites with MDX, versioning, i18n, and Algolia search built in.

Script Depot
◈Home🔍Search👤Me