use-local-whisper

Local voice transcription on macOS runs entirely on-device by switching from cloud APIs to a local whisper.cpp pipeline, generating transcripts without network dependency. This approach uses whisper-compatible models on Apple Silicon for enhanced privacy and zero latency.

whisper.cpp processes WhatsApp voice transcripts on Apple Silicon by executing local transcription pipelines. You must apply the voice-transcription skill first to ensure proper channel handling before switching transcription to the local whisper-cli binary.

On-device transcription with whisper.cpp requires installing whisper-cpp and ffmpeg, placing a GGML model in the data/models directory, and ensuring the whisper-cli binary is available. This setup enables local voice transcription on Apple Silicon Macs.

Local whisper transcription requires the voice-transcription skill to be applied first for proper WhatsApp channel handling. Without it, the on-device whisper.cpp pipeline cannot correctly process voice messages on macOS.

Switching to local whisper.cpp eliminates cloud API network usage and latency while generating transcripts entirely on-device. This approach enhances privacy by keeping audio data local on Apple Silicon Macs without transmitting recordings externally.