On-device voice-to-text that injects transcribed speech directly into any text field on your Mac. No cloud, no context switch.
You’re in VS Code. You just figured out why the caching layer is broken. The fix is clear in your head, but you also need to update your architecture notes before you forget the reasoning.
You could open your notes app, find the file, scroll to the right section, and start typing. Or you could just talk.
Dictation Built Into Strayfiles
The Strayfiles macOS app now includes voice dictation. Press Option+Space, speak, and your words appear wherever your cursor is — your editor, a markdown file, a chat window, even a terminal input.
No separate app. No browser tab. No cloud service. Speech recognition runs entirely on your Mac using Apple’s Speech framework.
How It Works
Hold Option+Space (or configure your own hotkey). A floating pill appears near your text cursor showing the recording state. Speak. Release the key. The pill shows “Processing” for a moment, then your transcribed text appears right where your cursor was.
That’s it. The text is injected directly into the focused application. Your clipboard stays untouched (unless the app requires it as a fallback).
Two Modes
Push-to-Talk works like a walkie-talkie. Hold the hotkey, speak, release. Best for quick thoughts and short dictations.
Toggle mode is for longer dictations. Press once to start, press again to stop. With auto-stop enabled, you don’t even need to press again — dictation ends automatically after 2 seconds of silence.
Toggle mode also has a hybrid behavior: hold the key for more than 300ms and it acts like push-to-talk. Quick tap keeps recording. One hotkey, two behaviors, no configuration needed.
Your Voice Stays on Your Machine
Most dictation tools route your audio through a cloud API. That means every word you say — including proprietary architecture decisions, internal API designs, and security-sensitive details — passes through someone else’s infrastructure.
Strayfiles dictation is fully local. The speech recognition model runs on your Mac. Audio is processed in real-time and immediately discarded. No network requests during transcription. No recordings saved to disk. No telemetry. No account required.
For developers working on proprietary codebases, this isn’t a nice-to-have — it’s a requirement. You can dictate notes about internal systems, document security decisions, and capture sensitive technical context without worrying about where that data ends up.
Models are downloaded once (75MB to 1.5GB depending on accuracy needs) and stored in ~/.strayfiles/models/. After that, dictation works entirely offline.
Smart Text Injection
The tricky part of voice dictation isn’t the transcription — it’s getting the text into the right place. Strayfiles uses a three-tier approach:
- Accessibility API — Reads the current text field, inserts at cursor, handles emoji and Unicode correctly
- Keyboard simulation — Types character by character for apps that don’t expose text fields
- Clipboard paste — Saves your clipboard, pastes, restores. Universal fallback.
The method is chosen automatically. It handles edge cases like emoji, surrogate pairs, and secure text fields without any configuration.
Dictation adds punctuation, capitalizes sentences, and strips filler words (“um”, “uh”, “like”) by default. All configurable per-app — you might want punctuation in your notes but not in your terminal.
Get Started
Open the Strayfiles macOS app, go to Settings > Dictation, and enable it. Grant Microphone and Accessibility permissions when prompted. The default hotkey is Option+Space.
Read the full documentation at strayfiles.com/docs/guides/dictation.