Who is it for?
- Useful when you want hold-to-talk dictation on an Apple Silicon Mac with speech recognition running locally.
- Useful when you also want local transcripts of meetings that capture your microphone and other participants' system audio.
Muesli — Audio to Text
Muesli-HQ
Muesli is an open-source Mac app that transcribes dictation on the device and pastes it into the active app, and can also transcribe meetings.
Why it fits
| Good to know | Desktop |
|---|---|
| Free output | Yes |
| No account needed | — |
| No watermark | — |
| Works offline | Yes |
| Batch processing | — |
| Content language | — |
— Unverified conditions. Official documentation reviewed. The app itself has not been hands-on tested. Matches only documented content-language support, not interface language.
Sources and review date4
Founder feedback and correction
Editorial selection guide
Last checked:
Getting started
- Install it with Homebrew or the DMG from the download page, keep the app in Applications and allow the macOS permissions it asks for.
- Hold the dictation shortcut, speak, and let the text paste into the app where your cursor is.
- For meetings, record or import an audio file, then check the transcript and notes before sharing them.
Decision points
- Choose it when open-source, Mac-only local dictation matters more than cross-platform availability.
- Meeting notes need a model you connect, such as an API-key provider or local Ollama, so compare that setup with all-in-one meeting note takers.
Limits
- The repository lists macOS 14.2 or later and describes the app as optimised for Apple Silicon; the site shows iOS only as coming soon.
- Permissions can include microphone, system audio recording, accessibility and input monitoring; models may download on first use.
- Notes generated through a connected provider can send the needed input to that provider; check the provider settings before using private recordings.
Frequently asked questions
Does Muesli work without internet?
The site says it works without internet after install and the docs say speech recognition runs on your Mac, though models may need to download on first use.
Can Muesli summarise meetings by itself?
Notes are generated by a model you connect: the README lists API-key providers, a ChatGPT sign-in, a local Claude Code CLI and local Ollama, so setup is required.