Transcribe.cpp – ggml based transcription engine

- **Transcribe.cpp** is an open-source, C++ transcription engine built on **ggml** (a tensor/matrix computation library widely used for running large language models locally, e.g., whisper.cpp, llama.cpp). - It runs **OpenAI's Whisper** speech-to-text models entirely on-device (CPU, GPU, or hybrid) with no internet connection needed, much faster than the reference Python implementation — suitable for offline transcription from microphones or files. - The project exists in the ecosystem of **ggml-based inference engines** (like llama.cpp, stable-diffusion.cpp). These tools have become popular in the open-source AI community because they allow powerful models to run on consumer hardware (laptops, phones, Raspberry Pi) instead of requiring cloud APIs. - This matters for privacy (data stays local), cost (no API fees), latency (real-time or near-real-time), and accessibility (works on budget or offline devices).