murmur

KO | EN

gitlite — search

#javascript #hacktoberfest #golang #go #machine-learning #rust #deep-learning #macos #windows #awesome #cli #awesome-list #vue #ios

(★ 14)

Lean, private, native-Swift speech-to-text for macOS: dictation, meeting capture, and file import, fully on-device.

murmur Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Murmur

**Free, private, native-Swift speech-to-text for macOS.**

Dictate into any app, capture meetings, and transcribe audio files, all on-device. Nothing is sent to the cloud.

### [⬇ Download for macOS](https://github.com/DerMaxl/murmur/releases/latest/download/Murmur.dmg)

<br>

---

## What it does

Murmur turns speech into text entirely on your Mac. Hold a key and talk, and what you say is typed at your cursor in whatever app you are using. It can also record meetings into clean, speaker-labelled transcripts and transcribe audio files you drop in. Keep it in the menu bar, the Dock, or both.

## Why I built it

I kept reaching for speech-to-text because talking is a more natural way to communicate, and to think, than typing. It quickly became how I worked with AI agents every day. But the tools I tried each fell short. One showed the words as they appeared, so if it stumbled or cut off I could carry on from where it left off, yet the transcription itself was mediocre. Another had a great model but hid the text while it formed, so a crash in the middle meant the whole message was simply gone. The polished apps that got both right sat behind paywalls that were hard to justify when the speech model is free and the app is simple to build.

So Murmur is built around one rule above all: your recording is always saved, even if something crashes. Around that sits a strong on-device model, no paywall, and nothing ever leaving your Mac.

## Features

- **Dictation anywhere:** hold a key, speak, and the text is typed at your cursor. Four interaction modes, a configurable hotkey, and your clipboard is preserved.
- **Transcribes while you talk:** long dictations are transcribed in the background as you speak, so the text lands almost instantly when you finish. Turn on the live preview to also watch the words appear as you go.
- **Mutes background audio** (Spotify, Music) while you talk and restores it after. Calls in a browser or an app like Zoom are left alone.
- **Meeting capture:** records your mic and the system audio as two tracks, with speaker labels merged into one chronological transcript.
- **File import:** drop in a voice memo or recording to transcribe it.
- **Never loses a recording:** audio streams to disk continuously and is recovered after a crash.
- **History and housekeeping:** searchable history, soft delete with a 30-day Recently Deleted, and auto-delete of old recordings.
- **Optional on-device AI:** a one-line summary and stutter cleanup per transcript.

## Speech engine and privacy

On macOS 26, Murmur starts on Apple's built-in speech engine: nothing to download, and dictation works the moment you install. For tougher audio or mixed languages, switch to NVIDIA Parakeet TDT v3 (via [FluidAudio](https://github.com/FluidInference/FluidAudio), on the Apple Neural Engine) in Settings — it covers German, English, Dutch, and 22 other European languages with automatic language detection, downloaded once on first use. On macOS 15, Parakeet is the engine. Everything runs on-device; the only network access is the model download and the update check. The optional AI features use Apple's on-device Foundation Models. To stay light, the speech models are unloaded after a few idle minutes and reloaded in seconds when needed (you can turn this off).

## Requirements

- Apple Silicon Mac, macOS 15 or later.
- The optional AI summary and cleanup need macOS 26 (skipped automatically otherwise).

## Download

[**Download Murmur**](https://github.com/DerMaxl/murmur/releases/latest/download/Murmur.dmg) (always the latest version), open the downloaded `Murmur.dmg`, and drag Murmur onto the Applications folder. It updates itself automatically from then on. The [releases page](https://github.com/DerMaxl/murmur/releases/latest) has the full notes.

Murmur is not notarized (it is a free personal project with no paid Apple Developer account), so macOS blocks the first launch with "Apple could not verify Murmur is free of malware." To open it the first time:

1. Open Murmur from your Applications folder. macOS shows the warning, click Done.
2. Open **System Settings > Privacy & Security** and scroll down to the Security section.
3. Next to "Murmur.app was blocked", click **Open Anyway**, then confirm with Touch ID or your password.

You only need to do this once. (On older macOS, right-clicking the app and choosing Open also works, but on macOS 15 the steps above are required.)

The first time you enable dictation or record a meeting, macOS asks for the relevant permission (Accessibility, microphone, or system audio). Grant it and you are set.

## For developers

Building from source, the design, and the roadmap are documented in [ARCHITECTURE.md](ARCHITECTURE.md) and [PLAN.md](PLAN.md).

A personal project, under active development.

murmur

murmur Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

murmur

murmur Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README