agentskillexchange Whisper.cpp Local Transcription Engine
Runs OpenAI Whisper models locally via whisper.cpp with GGML quantized weights for CPU-efficient transcription. Supports beam search decoding, VAD-based segmentation, and SRT/VTT subtitle output formats.
Installation
Use the upstream whisper.cpp build path for local transcription:
git clone https://github.com/ggml-org/whisper.cpp.git
cd whisper.cpp
sh ./models/download-ggml-model.sh base.en
cmake -B build
cmake --build build -j --config Release
./build/bin/whisper-cli -f samples/jfk.wav
For a quick demo, upstream also supports:
make base.en
The whisper-cli example currently expects 16-bit WAV input. Convert other audio formats before transcription when needed:
ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav
./build/bin/whisper-cli -f output.wav
Run ./build/bin/whisper-cli -h for detailed CLI options.
- Source: https://github.com/ggml-org/whisper.cpp
- Extracted from upstream docs: https://raw.githubusercontent.com/ggml-org/whisper.cpp/master/README.md