Voice to text for Mac — fully local, no cloud
evoglyph transcribes your speech and edits the result entirely on your device using the Apple Neural Engine. No audio leaves your Mac. No subscription.
Download for MacRequires Apple Silicon (M1 or later) · macOS 14 Sonoma or later
What is voice to text on Mac?
Voice to text (also called speech-to-text or dictation) lets you speak naturally and have your words appear as typed text in any app on your Mac. You activate it with a hotkey, speak, and the text lands wherever your cursor is — in an email, a document, a chat, a code editor, or a terminal.
Modern voice to text on Mac goes beyond simple transcription. Tools like evoglyph add an AI editing pass that automatically removes filler words, fixes punctuation, and formats your text — so what you get back is clean, formatted text, not a raw transcript.
The key architectural question for any voice to text tool is where the processing happens: on your Mac, or on a server somewhere. Cloud-based tools send your audio to remote servers for transcription. On-device tools run the speech model locally, which means faster round-trips and no audio leaving your machine.
How evoglyph works
evoglyph's pipeline has three steps, all running on your Mac with no network involvement:
-
Press your hotkey and speak
Press the configurable hotkey (default: the Fn/Globe key), speak naturally, then release. evoglyph records from your Mac's microphone directly — nothing is streamed to a server.
-
Local transcription and AI editing on the Apple Neural Engine
The audio is transcribed on-device by a Parakeet speech model running on the Apple Neural Engine. The raw transcript then goes through an in-process AI editing pass (a 1.7B-parameter model, quantized to run fast on Apple Silicon) that removes filler words, adds punctuation, and reformats the text. Both steps happen entirely on your device — no round-trip to the cloud at transcription or editing time.
-
Text injection into the active app
The clean, formatted text is injected directly into whatever app has focus. evoglyph uses three injection strategies depending on the app: AXUIElement splice for native Mac apps, keystroke injection for terminal apps, and pasteboard insertion for browser and Electron apps. You configure which strategy each app uses via the text injection settings.
evoglyph vs Apple Dictation
Apple Dictation is built into macOS at no extra cost. Here is how it compares to evoglyph on the points that matter most for daily use:
| Feature | evoglyph | Apple Dictation |
|---|---|---|
| On-device on Apple Silicon | Yes 1 | Yes, on Apple Silicon 2 |
| Automatic punctuation and editing | Yes — AI editing pass removes filler words, adds punctuation, reformats text automatically 1 | Auto-punctuation in supported languages; otherwise requires spoken commands (say "period", "new line") 2 |
| Configurable per-app text injection | Yes — three injection strategies configurable per app (AXUIElement, keystroke, pasteboard) 1 | Standard system text insertion; no per-app configuration |
| Editing mode customization | Yes — editable AI cleanup prompt, per-app modes | No AI editing pass; transcription only |
| Price | $20 one-time purchase; 3,000-word or 7-day free trial | Free (built into macOS) |
1 evoglyph architecture: Where your audio goes.
2 Apple Dictation on-device processing on Apple Silicon: Apple support page. Retrieved 2026-06-06.
Frequently asked questions
Try evoglyph free
3,000 words or 7-day free trial, no account required. $20 one-time purchase if you decide to keep it.
Download for MacApple Silicon (M1 or later) · macOS 14+