Vox
Show HN: I built free VoiceToText app with local LLM polish to save 60 min/day
Last verified:
What is Vox?
Vox is on-device AI dictation for Mac and Windows that lets users hold a key, speak, and paste cleaned-up text without any cloud round-trip. The app runs Whisper or NVIDIA Parakeet for transcription and uses Apple Intelligence (macOS 26+) or Gemma 4 for cleanup, with everything processing locally on your device.
Key features include hotkey-based dictation (⌘⌥. on Mac, Ctrl+Alt+. on Windows), five voice modes (General, Mail, Chat, Code Comment, Notes) that auto-tune cleanup based on the app you're typing into, and the ability to create custom modes with your own system prompts and post-processing rules. Vox works offline after the one-time model download on first run, requires no app account, and doesn't collect audio, transcripts, or crash reports.
Vox is designed for knowledge workers who type 3,000+ words daily at work—developers, PMs, designers, support staff, and anyone who Communicates via Slack, email, or documentation. The tool is also ideal for privacy-conscious users, people working on planes or without internet, and teams needing MDM-friendly deployment with central configuration.
Vox pricing
Pricing model: Freemium
Vox is fair-source: free for personal use (your own writing, side projects, hobby work) under the perpetual Personal Use license. Commercial licenses are required for use at work at companies with more than one person. Pricing starts at $12 USD/seat/month or $120/seat/year for 1-10 seats, with 10% off ($10.80/seat/mo or $108/seat/year) for 11-50 seats, and custom volume pricing for 51+ seats. Annual billing saves about 17% versus monthly. Stripe issues tax-compliant invoices for every charge, and PO-quoted invoices are available for 11+ seats.
Vox pros
- On-device processing—audio never leaves your device
- No cloud round-trip required at runtime
- No app account needed for personal use
- Works offline after first-run model download
- Hotkey dictation with ⌘⌥. on Mac and Ctrl+Alt+. on Windows
- Five voice modes tuned for General, Mail, Chat, Code Comment, and Notes
- Custom modes with your own system prompts and post-processing rules
- Auto-triggers per app (Slack, Gmail, VS Code, Notion, etc.)
- Cleans filler words and self-corrections automatically
- Enumerates lists and formats bullets on enumeration
- Preserves identifiers verbatim in code comments
- No dictation telemetry—audio, transcripts never written to disk
- Compatible with Apple Silicon M1+ and DirectX 12 GPU on Windows
- Fair-source license—free for personal use indefinitely
- Commercial license starts at $12/seat/month
- MDM-friendly installer for Jamf, Mosyle, Kandji, and Intune
- Central configuration sets team-wide default mode and dictionary
- Priority support with direct email line and one-day response
- Tax-compliant invoices from Stripe for every charge
- Annual plan saves ~17% versus monthly billing
- Estimated 3× faster than typing—153 WPM vs 52 WPM
Vox cons
- Intel Macs are not supported—Apple Silicon only
- Requires macOS 14+ or Windows 10/11 (x64)
- One-time model download needed on first run
- Does not auto-paste—user must press ⌘V or Ctrl+V
- Apple Intelligence cleanup is macOS-only
- Cleanup engine currently tuned primarily for English
- Parakeet is English-only (multilingual variants on roadmap)
- No SSO/SAML integration—local software only
Frequently asked questions about Vox
Does Vox work offline?
Yes. Audio is transcribed by a local model—Whisper or NVIDIA Parakeet—running on the Apple Neural Engine or Metal GPU on Mac, or any DirectX 12 GPU on Windows (with CPU fallback on both). Cleanup runs on Apple Intelligence (macOS 26+) or a local Gemma 4 model. No part of the pipeline requires internet at runtime after the first-run model download.
What gets sent to a server?
Nothing during dictation. Vox does not collect audio, transcripts, telemetry, crash reports, or analytics. The only network calls Vox makes are one-time model downloads on first run and an optional update check.
What hardware does Vox need?
On Mac: any Apple Silicon Mac (M1 or newer, late 2020 onwards) running macOS 14 or newer. Intel Macs are not supported. On Windows: 64-bit Windows 10 or 11 (x86_64). Any DirectX 12 GPU accelerates Parakeet transcription; CPU fallback works on older boxes. Gemma 4 cleanup runs on either platform; Apple Intelligence cleanup is macOS-only.
Do I need to set up a hotkey?
No. Vox ships with ⌘⌥. on Mac (Ctrl+Alt+. on Windows) bound by default, and it works without macOS Accessibility permission. You can rebind the hotkey in Settings if you prefer a different key combination.
Will Vox auto-paste text?
No. Vox copies the cleaned-up text to your clipboard, and you press ⌘V (Mac) or Ctrl+V (Windows) where you want it. This is deliberate: silent keystroke synthesis is fragile, requires Accessibility permission, and breaks confirmations in apps that intercept paste.
Is Vox open source?
The dictation pipeline ships with the desktop app, but you can verify what does and doesn't leave your machine with any network monitor like Little Snitch or LuLu on macOS, or GlassWire on Windows. The company is considering open-sourcing the modes engine and has asked users to let them know if it would matter.
Can I use Vox at work?
Vox is free for personal use—your own writing, side projects, hobby work, and personal email. If you use Vox as part of your job at a company with more than one person, you need a commercial license. The Teams plan covers commercial use, MDM-friendly deployment, and central configuration.
What about other languages?
Vox supports Whisper (multilingual) and Parakeet (English, with multilingual variants on the roadmap). The cleanup engine is currently tuned for English; other languages will produce competent transcripts with light cleanup until per-language tuning ships.
What does the commercial license grant?
Per-seat permission for one named employee to use Vox as part of their work for the licensed organization, including on company-owned hardware. Personal use of Vox by the same employee outside of work doesn't consume an additional seat. The license is reissued automatically as your seat count changes.
How is Vox different from cloud dictation tools?
Most dictation tools send audio to a server; Vox processes everything on your device. Vox requires no internet at runtime, no app account, never retains audio after transcription, has no dictation telemetry, works on a plane, and is independently verifiable with network monitors. Cloud tools typically require internet, accounts, and vary by vendor in retention and telemetry.