# Muesli

Muesli is an open-source Mac app for on-device dictation, local-first meeting transcription, and private speech-to-text on Apple Silicon.

## Primary Pages
- Product overview: https://muesli.works/
- On-device Dictation: https://muesli.works/on-device-dictation - Mac dictation that runs speech-to-text locally on Apple Silicon and pastes clean text into the app you are already using.
- Meeting Notes: https://muesli.works/meeting-notes - AI meeting notes without a meeting bot, using local transcription from your Mac microphone and system audio.
- Local-first AI: https://muesli.works/local-first-ai - A local-first AI Mac app where speech-to-text starts on-device and optional cloud providers stay explicit.
- Mac Dictation App: https://muesli.works/mac-dictation-app - A practical guide to Mac dictation apps, local speech-to-text, offline models such as Parakeet and Whisper, and private voice typing on Apple Silicon.
- Best Dictation Apps for Mac: https://muesli.works/best-dictation-apps-mac - A fair 2026 guide to the best dictation apps for Mac and macOS, including Apple Dictation, Muesli, Superwhisper, Wispr Flow, VoiceInk, and other voice-to-text options.
- Offline Dictation for Mac: https://muesli.works/offline-dictation-mac - A practical guide to offline dictation on Mac, local speech-to-text models, and what can run without sending audio to a cloud transcription service.
- Apple Neural Engine Speech-to-Text on Mac: https://muesli.works/apple-neural-engine-speech-to-text-mac - A technical guide to Apple Neural Engine speech-to-text on Mac, CoreML, local ASR models such as Parakeet and Whisper, and how Muesli uses Apple Silicon for local dictation.
- Local Speech-to-Text Glossary: https://muesli.works/local-speech-to-text-glossary - A technical glossary for local speech-to-text on Mac, covering ASR versus speech-to-text, CoreML, Apple Neural Engine, Parakeet, Whisper, Qwen3 ASR, VAD, diarization, neural AEC, LocalVQE, and local-first transcription.
- Local Meeting Transcription for Mac: https://muesli.works/local-meeting-transcription-mac - A practical guide to local meeting transcription on Mac and macOS, meeting notes without a bot, and what can stay on your own computer.
- Granola Alternative: https://muesli.works/granola-alternative - A practical Granola alternative guide for people who want local-first meeting notes, open-source software, and more ownership of their workday memory.
- Wispr Flow Alternative: https://muesli.works/wispr-flow-alternative - A practical Wispr Flow alternative guide for Mac and macOS users who want local-first dictation, offline speech models, and more ownership of voice-to-text.
- Otter.ai Alternative: https://muesli.works/otter-ai-alternative - A practical Otter.ai alternative guide for macOS users who want local-first meeting transcription, no meeting bot by default, and notes they can own.
- Fireflies.ai Alternative: https://muesli.works/fireflies-ai-alternative - A practical Fireflies.ai alternative guide for macOS users who want meeting transcription to start on their Mac without sending a bot into every room.
- Help: https://muesli.works/help
- Changelog: https://muesli.works/changelog
- GitHub: https://github.com/pHequals7/muesli

## Key Facts
- Runs speech-to-text locally on Apple Silicon.
- Supports hold-to-talk dictation and meeting transcription.
- Captures meeting audio from your own Mac without adding a meeting bot.
- Stores dictations, transcripts, and meeting notes locally by default.
- Open-source and inspectable on GitHub.
- Optional services such as OpenAI, OpenRouter, ChatGPT, and Google Calendar are explicit integrations, not the default transcription path.

## Technical Facts
- Muesli is a native macOS app built with Swift and AppKit, not Electron.
- Local speech-to-text runs on Apple Silicon through CoreML and Apple Neural Engine-capable model paths where supported.
- For short dictation, local Apple Silicon inference can reduce latency by avoiding the cloud upload, remote inference, response, and paste-back round trip.
- The Apple Neural Engine is dedicated neural network hardware, which helps supported speech models run efficiently on device.
- Muesli supports local ASR model options including NVIDIA Parakeet, Whisper, Qwen3 ASR, and Nemotron Streaming.
- The default dictation workflow does not require a hosted speech-to-text API after local models are installed.
- Meeting transcription combines microphone audio, system audio capture, voice activity detection, and speaker diarization.
- Meeting echo cancellation runs locally with bundled LocalVQE by default and DTLN available as a fallback AEC path.
- Parakeet and Nemotron come from NVIDIA, Whisper comes from OpenAI, Qwen3 ASR comes from Alibaba’s Qwen model family, and Cohere Transcribe comes from Cohere.
- Optional summarization providers are separate from the local transcription path.

## Social Links
- https://github.com/pHequals7/muesli
- https://x.com/fastspeech2text
- https://www.linkedin.com/company/mueslios/