🛡

100% On-Device
Processing

Unlike cloud-based transcription services that upload your audio to remote servers, MetaWhisp processes everything locally using Apple's Neural Engine and the WhisperKit framework. Your voice data stays on your hardware at all times.

  • Zero network requests during transcription
  • Audio is discarded immediately after processing
  • No telemetry, no analytics, no tracking
  • Works perfectly in airplane mode
  • OpenAI key optional — only for post-processing

Zero Network Activity

During transcription, MetaWhisp makes exactly zero network requests. You can verify this yourself using any network monitoring tool. The entire Whisper model runs on-device through Apple's Core ML framework, using the Neural Engine on Apple Silicon chips for hardware-accelerated inference.

The only time MetaWhisp accesses the network is during the initial one-time model download (~1.5 GB) and when you optionally use AI post-processing features that require an OpenAI API key.

Audio Is Never Stored

When you record speech, the audio data exists only in memory for the duration of processing. The moment transcription is complete, the audio buffer is discarded. No audio files are written to disk, no recordings are cached, and no voice data persists after the transcription is done.

No Telemetry, No Analytics

MetaWhisp contains no analytics SDKs, no crash reporting services, no usage tracking, and no phone-home mechanisms. We do not collect any data about how you use the app, what you transcribe, or when you use it. There are no anonymous usage statistics, no opt-in telemetry, and no hidden data collection.

Airplane Mode Ready

Once the AI model is downloaded, MetaWhisp works with no internet connection at all. Turn on airplane mode, disconnect from Wi-Fi, or work in a secure facility — transcription continues working at full speed and accuracy. This is ideal for professionals working with sensitive information in restricted environments.

OpenAI Key Is Optional

The AI post-processing features (grammar correction, rewriting, and translation) use OpenAI's GPT API. These features are entirely optional. If you choose to use them, you provide your own API key, and the only data sent to OpenAI is the transcribed text (not your audio). If you never configure an API key, MetaWhisp never makes any network requests after the initial model download.

Open Architecture

MetaWhisp uses the open-source WhisperKit framework by Argmax for on-device inference. The Whisper large-v3-turbo model from OpenAI is well-documented and widely audited. There are no proprietary black boxes in the transcription pipeline.

Privacy without compromise

Experience transcription that respects your data. Download MetaWhisp today.

Download for macOS

macOS 14+ · Apple Silicon · Free