whisper.cpp: offline voice transcription on CPU with pure C++

OpenAI had already released Whisper, a powerful model for transcribing audio to text. The catch? You needed Python, CUDA, and a decent GPU. Not exactly accessible to everyone.

Georgi Gerganov — the same developer who would soon do the same thing for LLMs — rewrote Whisper from scratch in pure C++. The result is a lightweight program that runs on any CPU, even an old laptop, with no internet connection.

In practice: you can transcribe meetings, voice notes, interviews, or podcasts directly on your computer, with very high quality, without sending anything to any server.