Default
On-device Apple Speech
macOS' native dedicated ASR runs first — fast, offline, no cloud round-trip. ~95% of dictations never leave your Mac.
PromptPilot Desktop · macOS · Apple Silicon
Speak. Prompt. Ship.
PromptPilot Desktop — dictate with Voice, open Text and Image Studio, and manage your account from one native tray app. Hold a hotkey, talk, release — and a polished transcript lands at your cursor.
macOS · Apple Silicon
Quick install
$ curl -fsSL https://promptpilotai.in/install-voice.sh | bashListening· hold to dictate
⌘⇧Space
ChatGPT
draft a quick note about the deployment plan for tomorrow
Why this exists
Dictation apps give you a transcript. PromptPilot Desktop gives you Voice — stutters, fillers, and dropped punctuation already fixed — pasted right where your cursor was. Plus Text and Image Studio from the same tray app — you never leave your flow.
How it works
Four steps. About a second of latency.
Groq-hosted Whisper Large v3 handles transcription by default — faster and more accurate than on-device recognition. Users who prefer on-device speech can flip the default in Settings.
01
Hold the hotkey
⌘⇧Space, ⌥Space, or Fn. Double-tap to latch hands-free.
02
Speak naturally
Groq-hosted Whisper transcribes by default. Fully on-device mode available in Settings.
03
AI cleans the transcript
A structured-output pass strips fillers and fixes punctuation. Meaning preserved verbatim.
04
Paste at your cursor
Lands in Notes, VS Code, Slack, ChatGPT — wherever your cursor is.
Features
Built for the way you actually type.
Every detail tuned for fast, faithful dictation — not for a demo reel.
Double-tap to latch
Tap the hotkey twice for hands-free recording. Tap once to stop. Long meetings, long thoughts, no held-down keys.
Raw mode for verbatim
⌘⇧R skips the cleanup pass — your words paste exactly as spoken. For code, names, quotes.
Brand-grade
Structured AI cleanup
Gemini Flash Lite via Vercel AI Gateway. Zod-constrained output means it cannot refuse, paraphrase, or add commentary.
Paste-anywhere
Native CGEventPost drops text at your cursor in any focused app — Notes, VS Code, Slack, ChatGPT, anywhere.
Menu-bar native
Lives in the menu bar. No dock clutter, no window-switching. Transparent always-on-top HUD pill while you speak.
Searchable history
Every session lands in your PromptPilot history. Copy, re-use, or feed it back into the web optimizer.
Auto-updates
Sparkle-style updater wired in — new builds land in the background, no App Store, no friction.
One PromptPilot
Your account. Your history. One plan.
Desktop is not a separate product with a separate bill. Voice dictation, Studios, and your web account share one PromptPilot plan — sign in once and every Voice session lands in the same history you already use on the web.
Part of the PromptPilot family — not a side product.
Desktop was built against the same Next.js app, the same Supabase schema, the same AI Gateway routing layer. Switching from keyboard prompting on the web to Voice dictation — or opening Text and Image Studio from the tray — is one keypress, not a context switch.
- AccountSign in once. Voice shares your PromptPilot session via secure desktop handoff.
- PlanVoice uses your existing PromptPilot plan — no separate subscription, no second bill.
- ModelsCleanup and ASR fallback are routed via Vercel AI Gateway — admin-tunable in model_configs.
- HistoryEvery voice session lands in your PromptPilot history. Searchable, replayable, exportable.
System requirements
What you'll need.
macOS
14 Sonoma or later
Chip
Apple Silicon (M1 and up)
Microphone
Required — built-in or external
Speech Recognition
Required — for on-device ASR
Accessibility
Required — to paste at your cursor
Windows
Coming in v0.2
Stop typing the obvious stuff.
Hold ⌘⇧Space or ⌥Space anywhere on your Mac, watch a clean prompt land in the app you were already using.
macOS · Apple Silicon
Quick install
$ curl -fsSL https://promptpilotai.in/install-voice.sh | bash