Can I vibe code Superwhisper?
On-device voice dictation that types anywhere on your Mac.
Whisper-class models run locally now, and the app around them — hotkey, record, transcribe, paste into the focused field — is a scoped native utility an agent can genuinely build. Riley Brown vibe-codes with voice tools like this daily; owning yours means your dictation style prompts, your vocabulary, your rules. The paid apps win on polish (instant model switching, context modes), but the core loop is a weekend of Swift you'll use forever.
Build a macOS dictation utility in Swift/SwiftUI. Menu-bar app with a global hotkey: press to record mic audio, release (or press again) to stop; transcribe locally with whisper.cpp (bundle a small model, allow pointing at a larger one); then an optional cleanup pass through an LLM API applying my rules from a style.md (capitalization, filler-word removal, 'new line' commands); finally paste the text into the frontmost app via accessibility APIs, with clipboard fallback. Menu shows last transcription and a toggle for the cleanup pass. Document mic + accessibility permissions in the README.
A Mac menu-bar dictation app: global hotkey records, local Whisper transcribes, optional LLM cleanup pass applies your style rules, result pastes into whatever field has focus.
- Polished mode system (email mode, code mode, prompt mode)
- Tuned latency — theirs feels instant, yours will feel fast-ish
- Automatic app-aware context and formatting
- Zero-setup installer your less technical friends could use
Dictation is a hundred-times-a-day tool; every 200ms of latency and every formatting miss compounds. The paid apps have sanded those edges for years.
A prompt is the first move, not the whole game. Course 01 teaches you to take a prompt like this one to a shipped, working app — reviewing, correcting, and steering the agent the whole way.
Start Course 01 →