MetaWhisp was designed from the ground up with privacy as a core principle, not an afterthought.
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.
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.
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.
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.
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.
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.
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.
Experience transcription that respects your data. Download MetaWhisp today.
Download for macOSmacOS 14+ · Apple Silicon · Free