01
Hold Control + Option
HeyVedu listens only while you hold the shortcut.
Private dictation for Mac
Free, open-source dictation that turns natural speech into polished text anywhere on macOS—on your device, even when you’re offline.
Free foreverNo accountNo historyOn-device
One shortcut. Any text field.
HeyVedu stays out of your way and does one focused job clearly.
01
HeyVedu listens only while you hold the shortcut.
02
Pause, use filler words, or correct yourself mid-sentence.
03
Clean text appears wherever your cursor is.
Privacy, by architecture
HeyVedu’s default dictation path runs on your Mac. There is no HeyVedu cloud, account, or transcript archive.
Clean, not rewritten
HeyVedu removes the verbal clutter while preserving what you meant, how you meant it.
“Let’s meet Friday—actually, Monday afternoon.”
Let’s meet Monday afternoon.
Meaning preserved“Hey, um, could you send me the, uh, revised proposal?”
Hey, could you send me the revised proposal?
Fillers removed“Three things for tomorrow: finish the brief, email Sam, and book the room.”
Small app. Clear principles.
Speech recognition and standard cleanup happen on-device.
No history, no audio files, and no HeyVedu account.
GPLv3 source that people can inspect, modify, and contribute to.
Messages, notes, email, editors, terminals, and most other text fields.
Transparent optional modes
Standard transcription and cleanup stay local. Advanced users can optionally choose a locally installed and authenticated Claude Code or Codex CLI for cleanup.
When selected, transcript text and your configured vocabulary are sent through your existing provider account. Audio and speech recognition remain on-device. These modes are not offline.
Changed anytime from the menu bar.
Open, all the way down
HeyVedu is GPLv3-licensed and built as a focused native macOS menu-bar utility. Its privacy model isn’t a promise hidden in a policy—it’s visible in the source.
// Keep the clipboard exactly as we found it.
let saved = snapshot(pasteboard)
item.setString(text, forType: .string)
item.setData(Data(), forType:
transientType)
postPaste()
// Restore it after the text is inserted.
Before you begin
HeyVedu is currently installed from source. There is no signed binary release yet—we’d rather say that plainly.
./scripts/run.sh1 Clone or download the repository
2 Run the command from the project folder
3 Grant Microphone and Accessibility access
Source builds are ad-hoc signed, so macOS may ask for Accessibility access again after rebuilding.
Good to know
Yes. HeyVedu is free and open source under the GNU GPLv3. There is no HeyVedu account or subscription.
Yes, after the initial build and model downloads. The default speech recognition and cleanup pipeline runs entirely on your Mac. Optional Claude Code and Codex CLI cleanup modes require their corresponding online provider.
No. Recorded audio stays in memory, is never written to disk, and is transcribed on-device—even if you choose an optional advanced cleanup engine.
No. HeyVedu does not store a dictation history. Debug builds log timing and errors, but not audio, transcripts, vocabulary, or typed keys.
Accessibility access lets HeyVedu observe the Control + Option push-to-talk shortcut and paste the finished text at your current cursor.
HeyVedu requires an Apple-silicon Mac running macOS 26.4 or later. Dictation is currently English-only.
Clone or download the GitHub repository, then run ./scripts/run.sh. You’ll need Xcode with the macOS 26.4 SDK. There is no signed binary release yet.
They are optional cleanup modes that use your locally installed and authenticated Claude Code or Codex CLI. Transcript text and configured vocabulary may be sent to Anthropic or OpenAI through your existing account; audio stays local.
Dictate on your terms
Turn your voice into polished text without turning it into someone else’s data.