About
Why we built Breuzel
We were heavy Wispr Flow users. Then we read the privacy policy.
Every voice dictation tool we tried had the same problem: they route your audio through their own servers, store it, and charge you a flat subscription that hides the actual cost. Some use it to train their models. None of them let you use your own API key.
With Groq's Whisper API now available at $0.111/hour and sub-second latency, the math changed completely. Why pay $15/month to a middleman when you can pay Groq directly at $0.0018/minute — with your own key that never leaves your Mac?
Breuzel is the tool we wanted to exist. Mac-native, BYOK, fast as hell, and honest about every cent you spend.
What we believe
Privacy is structural, not a checkbox
We don't just say we care about privacy. We built Breuzel so that we are architecturally incapable of accessing your data. Your key is in Keychain. Your audio goes directly to Groq. There's nothing for us to leak.
Cost should be transparent
Every dictation session shows your exact API cost. Not an estimate. The actual Groq charge, in dollars and cents, live in your menu bar. You should always know what you're spending.
Speed is a feature
A tool that takes 3 seconds to transcribe gets abandoned. We obsessed over the end-to-end pipeline to hit sub-1500ms from hotkey to injected text. It feels instant because it nearly is.
Native beats web for this use case
You cannot build a global hotkey or write to another app's text cursor in a browser. We built Breuzel as a Tauri app (Rust + React) so it has full macOS Accessibility API access. No Electron bloat.
Changelog
- +Initial release
- +Groq Whisper large-v3 transcription
- +Claude Haiku post-processing
- +macOS Keychain BYOK storage
- +Global hotkey (⌃⇧Space)
- +Amber privacy badge
- +Session cost display
Roadmap
- Web dashboard (usage charts, billing)
- OpenAI Whisper BYOK support
- Email support channel
- Local Whisper mode (on-device, no API)
- Windows support (Windows Credential Manager)
- Custom dictionary for technical terms
- Voice commands (delete last sentence, etc.)
- HIPAA BAA (requires on-device processing)
- Teams workspace & shared usage dashboard
- AI transformations (rewrite as email, summarise)