whisper.cpp
Offline speech-to-text inference in C/C++
48.7k|stable|v1.8.4
Download for Windows
Download .zip — extract and run
How to install
1Download the file above
2Extract the .zip file
3Run whisper.cpp.exe
No external dependencies — Plain C/C++ implementation runs standalone on any platform.
Hardware acceleration — Optimized for Apple Silicon, x86 AVX, NVIDIA GPU, Vulkan, and CoreML.
Model quantization — Integer quantization reduces memory and disk usage while maintaining quality.
Cross-platform support — Runs on macOS, iOS, Android, Linux, Windows, WebAssembly, and Raspberry Pi.
Voice activity detection — Built-in VAD skips silent segments for faster transcription.
High-performance inference engine for OpenAI's Whisper with minimal dependencies and broad hardware support. Runs offline on Apple Silicon, x86, NVIDIA GPUs, and mobile devices with near real-time speed. Used in production for on-device transcription across platforms.
C++11 compiler; CMake for building; FFmpeg for audio conversion
Source on GitHub
pavilion