use-local-whisper

Run WhatsApp voice transcription entirely on-device by switching from OpenAI's Whisper API to local whisper.cpp, which requires modifying src/transcription.ts to use whisper-cli. This enables offline operation with no API keys or network costs.

Local whisper.cpp transcription requires macOS with Apple Silicon, whisper-cpp, ffmpeg, and a local GGML model. You must also ensure the Voice-Transcription skill is applied for WhatsApp before running npm run build.

Yes, on-device whisper transcription can be extended to other messaging platforms. It currently targets WhatsApp, but you can add channel-specific audio download logic to support other channels within your workflow.

Switch from the OpenAI Whisper API to local whisper.cpp to enable offline operation where no data leaves the device. This approach eliminates network dependencies and API costs while supporting privacy-focused workflows.

Yes, whisper.cpp transcription runs entirely offline without an internet connection. Once the local GGML model is downloaded, on-device processing handles voice notes independently without any network traffic or data leaving the device.