use-local-whisper

You can transcribe WhatsApp voice messages locally by applying the voice-transcription skill and then enabling on-device processing with whisper.cpp. This approach uses ffmpeg for audio conversion and whisper-cli for transcription to avoid API keys, network usage, and costs.

On-device transcription with whisper.cpp is the process of converting audio to text entirely on your local machine. It uses a GGML model and whisper-cli to process audio files without sending data to external servers or requiring network access.

Yes, local voice transcription with whisper.cpp supports both Apple Silicon and x86 environments. The transcription pipeline utilizes ffmpeg for audio conversion and whisper-cli, which is compatible with these architectures when configured properly.

Yes, you need ffmpeg for audio conversion and whisper-cli for transcription to process voice messages on your device. The system uses these tools alongside a GGML model, so they must be available in your environment.

You configure local transcription by setting the WHISPER_BIN and WHISPER_MODEL environment variables to point to your whisper-cli executable and GGML model. The system respects these variables during the transcription process.

Local voice transcription requires the voice-transcription skill first because it provides the foundational phase-based integration needed to detect and handle WhatsApp voice messages. The local whisper.cpp processing builds on top of this prerequisite setup.