Murmela
Private dictation for Mac
Your thoughts.
On your Mac.
Press a shortcut, speak, and your words appear in any app. Fast, accurate and completely offline. No cloud, no account, no one listening.
For Macs with Apple silicon · Say it: moor-MEH-la
Talk instead of type, wherever your cursor is.
Mail, Slack, Notes, Word, your browser or your code editor. If you can type there, you can talk there.
Press the shortcut
Hold it while you talk, or tap to start and stop. Pick any shortcut you like.
Just talk
Speak naturally. Punctuation and capitalization are handled for you.
Text appears
Let go, and your words land right where you were typing.
Private by design
Nothing leaves your Mac.
Most dictation tools send your audio to a server. Murmela runs the speech models on your Mac's own chip, so there's nothing to send.
Works fully offline
Download a speech model once, then turn off Wi‑Fi. It keeps working on a plane, on a train or in a locked-down office.
Nothing is uploaded
Your audio and your text are processed and stored only on your Mac. No servers, no tracking, no data to leak.
No account needed
No sign-up and no login. Open the app and start talking. Your history is yours to keep, search or switch off.
Cloud dictation
Your voice → the internet → a server → a company's logs → back to you.
Murmela
Your voice → your Mac → your text. That's the whole trip.
Everything you say, ready to use.
Transcribe audio and video files
Drop in interviews, meetings, voice memos or videos (MP3, M4A, WAV, MP4, MOV and more), up to six hours long.
Teach it your words
Add names, brands and jargon to your vocabulary, so “Murmela” never becomes “Marmalade” again.
A searchable library
Every transcript is saved on your Mac and fully searchable. Prefer no history? Turn it off with one switch.
Choose speed or accuracy
Pick from leading open speech models, from instant and lightweight to maximum accuracy for long recordings.
Translate as you speak Beta
Speak in English or German and paste in French or Spanish, still fully offline.
Made for Apple silicon.
M1 or newer
Any Mac with an M‑series chip runs speech models locally, fast.
macOS 11 or later
Big Sur and up. A few advanced models need macOS 14.
Online once
Internet is only needed to download a speech model. After that, never.
Good to know.
Is it really 100% offline?
Yes. The internet is used only to download the speech models you choose. After that, recording, transcription and translation all run on your Mac, and you can disconnect completely.
How is this different from the dictation built into macOS?
Murmela uses larger, more accurate speech models, handles long dictations, learns your custom vocabulary, transcribes audio and video files with speaker labels, and keeps a searchable library, all without sending anything to a server.
Does it work on Intel Macs, Windows or iPhone?
Murmela is built and tuned for Macs with Apple silicon (M1 and newer), which can run powerful speech models locally. Join the waitlist and tell us if you need another platform.
How much will it cost?
There will be a free version. Pricing for the full version will be announced at launch, and waitlist members hear about it first.
When does it launch?
Soon. Murmela is in final testing now. Waitlist members get access first.
Say it out loud.
Keep it to yourself.
Leave your email and we'll let you know the moment Murmela is ready to download.
No spam. Just one email when we launch.