Frequently asked questions

A macOS voice dictation app. You press a shortcut, you speak, and the text is written at the cursor in the active application. All recognition happens on your Mac.

Whisper models recognise around a hundred languages, with quality varying by language; DictateLocal takes particular care with French (punctuation, quotation marks, times). The optional, faster Parakeet engine covers 25 European languages. You can fix the language, combine two (“fr+en”) or leave detection automatic. The interface is translated into ten languages.

DictateLocal uses newer, more accurate models, works in every application with a single shortcut, and adds automatic French punctuation, per-app profiles, formatting modes, translation, subtitles and meeting mode. And it shows you exactly what leaves your Mac.

Whisper (through WhisperKit) in five variants, from the lightest to the most accurate, and Parakeet TDT v3 (through FluidAudio), roughly ten times faster. Each model is downloaded once from huggingface.co and then used offline. The first load on a given macOS version triggers a Core ML optimisation that can take over a minute.

Formatting applied to the dictated text: Fix, Clear and concise, Professional tone, Email, Bullet list, Summary, “Prompt for an AI” (turns your spoken request into a structured instruction) and “Code prompt” (ready for Claude Code, Cursor or Copilot). They rely on the macOS language model and run on the Mac. A shortcut opens a palette to switch modes from the keyboard.

“New line”, “new paragraph” and “undo that” (which removes the last insertion), plus their French equivalents. Filler-word cleanup (“um”, “uh”) is adjustable and never rewords your sentence.

Yes: open dictatelocal://stop, cancel, pause, settings, files. The start, toggle, subtitles and meeting commands require enabling “External commands” in Settings, General.

No. Audio is captured in memory, transcribed locally, then released. Neither audio nor text is sent to a server, and DictateLocal has no user account and no analytics.

Only three destinations: api.lemonsqueezy.com to activate and recheck the license, dictatelocal.com to look for updates, and huggingface.co to download models. If you yourself enable a cloud engine (Groq or OpenAI) with your own key, audio is sent to that provider, and only to it, after explicit consent. The app’s network log (Settings, About) lists every connection.

On your Mac, encrypted with a key kept in the Keychain. You can browse, clear or disable it in Settings, History.

Three days from first launch, every feature, no account or card. After that, the app asks for a key to keep dictating.

Your choice: a lifetime license at €79 (about $85) incl. VAT, paid once (launch price, instead of €99), or a subscription at €6.99/month (about $7.50) or €49/year (about $53), cancel any time. Either way: three Macs, every feature, updates included. Prices in EUR; your local currency is shown at checkout. Comparison on the Pricing page.

The subscription ends at the end of the paid period and the app becomes read-only: no more dictation, but your settings and history stay on your Mac. You can cancel from the Lemon Squeezy receipt email, and switch to the lifetime license whenever you like.

Settings, About: paste the key you received by email into “License key” and click “Activate”. To free a Mac: “Deactivate on this Mac”.

Look for the Lemon Squeezy email (subject mentioning “DictateLocal”), or write to [email protected] from the address used at purchase.

Yes, within 14 days of purchase, no questions asked. Email us; Lemon Squeezy refunds the original payment method.

macOS 26 (Tahoe) or later. A Mac with Apple silicon is recommended: recognition is much faster and the on-device AI modes rely on the macOS language model.

Download the disk image, open it and drag DictateLocal to Applications. The app is signed with a Developer ID certificate and notarized by Apple. On first launch it walks you through the permissions it needs.

Microphone (required), Accessibility (to insert text into other apps), Input Monitoring (only if you pick a single key as “hold to talk”) and Screen & System Audio Recording (only for live captions and meeting mode). Details on the Support page.

Quit the app, drag it to the Trash, then optionally delete ~/Library/Application Support/DictateLocal (models and history). The full procedure is on the Support page.

Once a day the app checks, through Sparkle, whether a new version is available and offers it to you. Updates are signed and verified before installation. You can also check manually from the menu bar menu.