use-local-whisper

You can transcribe WhatsApp voice messages locally by replacing the remote OpenAI Whisper API with a local whisper.cpp CLI workflow. This skill implements a deterministic transcription path that processes audio attachments entirely offline, avoiding API costs while preserving voice message metadata.

Yes, ffmpeg is required for audio format conversion in the local transcription pipeline. The skill uses ffmpeg to handle audio format conversion before passing the audio streams to the whisper.cpp CLI, ensuring compatibility with the on-device transcription workflow.

Yes, this skill supports on-device transcription locally on macOS with Apple Silicon using whisper.cpp. It operates entirely offline when dependencies are installed, requiring no API keys and avoiding external network usage during audio processing.

The best way to avoid API costs is to implement a local whisper.cpp CLI workflow instead of using remote transcription services. This skill replaces remote OpenAI Whisper transcription with an offline pipeline, exposing a stable interface while handling audio conversion and error management automatically.

Yes, the local whisper.cpp transcription pipeline preserves the isVoiceMessage flag and error-handling behavior. It exposes a stable transcribeAudioMessage interface that maintains the existing metadata structure while routing audio through the local CLI workflow.