Zero shenanigans.
Local and open source. A tool that listens to you should never report back about you.
Read the code.
hyprwhspr is fully open source under the MIT license. Every line that touches your microphone, audio, and keystrokes is public and auditable — nothing compiled, obfuscated, or hidden away.
Want to verify exactly what happens to your audio, from hotkey to text? You can.
Browse the source on GitHub ↗No company. No incentives.
hyprwhspr is built and maintained by a single developer. There's no company behind it, no investors, and no business model that depends on your data.
It exists because it's useful and building it was interesting. Nobody stands to gain from your data, so nobody ever will.
Local means local.
With a local backend — Parakeet, Whisper, onnx-asr, Cohere Transcribe — your audio is processed entirely on your own hardware. It is never sent to a server, never stored unless you ask it to be, and never leaves your machine.
There's no telemetry, analytics, or crash reporting, and during normal operation it makes no network calls at all. The service is a local systemd process with no reason to phone home and nowhere to phone home to.
A note on cloud backends
hyprwhspr optionally supports cloud transcription APIs — Google Gemini, OpenAI, Groq, Cohere, ElevenLabs, and any OpenAI-compatible endpoint. With a cloud backend, your audio is sent to that provider's servers and subject to their privacy policies.
These are strictly opt-in: the default installation uses local models only, and a cloud backend won't work until you configure an API key yourself. It's a deliberate choice you make — not something you're silently enrolled in.