Common speech-to-text architectures, explained without the fog
CTC, RNN-T, TDT, Conformer encoders, encoder-decoder Transformers, and the surrounding systems that make local ASR useful.
Muesli field notes
Speech-to-text should be something you can understand, run, and own.
Notes on local ASR, Mac dictation, meeting capture, privacy, model architecture, and the tools that shape how speech becomes part of your workday.
LOCAL-FIRST RESEARCHless cloud
more controlPractical writing for people who want to know where their transcript went.
more controlPractical writing for people who want to know where their transcript went.
The best medical dictation tool might start on your own Mac
A privacy-first look at medical dictation, clinical scribes, and the local speech-to-text layer doctors can control before notes enter an approved system.
Read the field noteTHE ARCHIVE
Useful things to know before you rent out your voice.
18 field notes
NVIDIA Parakeet makes local English speech-to-text feel practical
Why Parakeet matters for fast local English ASR, and why a clear sentence from your Mac should not need a cloud round trip.
OpenAI Whisper speech-to-text: where it still shines locally
A practical guide to Whisper’s encoder-decoder design, multilingual strengths, local inference tradeoffs, and where it fits in Muesli.
Granola has the sugar. Muesli keeps your transcript safe and healthy.
Why local speech-to-text has advanced far enough that every meeting transcript does not need to begin in someone else’s cloud.
The best offline dictation app for Mac should not treat local speech as an afterthought
A practical comparison of offline dictation apps for Mac, from built-in Apple Dictation to local model workflows on Apple Silicon.
Meeting notes without adding another guest to the call
What changes when your Mac captures the meeting instead of a bot joining Zoom, Meet, or Teams on your behalf.
Apple Dictation is a feature. Muesli is a speech workspace.
When built-in voice typing stops being enough: local model choice, transcript history, meeting transcription, and a workflow you control.
The local speech-to-text glossary for Mac
ASR, VAD, diarization, neural AEC, CoreML, Apple Neural Engine, Parakeet, Whisper, and the terms behind a local speech stack.
What Apple Silicon changes about speech-to-text on Mac
How CoreML and the Apple Neural Engine change the latency, power, and privacy equation for local dictation and transcription.
What should a Mac dictation app actually do?
A practical guide to local voice typing, paste-at-cursor workflows, model choice, privacy, and where Muesli fits.
How local meeting transcription works on Mac
Capture microphone and system audio from the Mac already in the call, then keep the transcript close before you decide what to share.
Offline dictation on Mac is finally a real everyday workflow
What offline means in practice, how local ASR models fit, and why the useful path is still speak, release, and get text where your cursor is.
A local Granola alternative for Mac meeting notes
For people who want Granola-style meeting memory without making a hosted workspace the default home for every transcript.
A Superwhisper alternative for people who want the speech layer on their Mac
A practical comparison of polished AI dictation and Muesli’s open-source local-first path for dictation and meeting transcription.
The best Wispr Flow alternatives for Mac in 2026
A practical comparison of Wispr Flow alternatives by platform, privacy, offline speech-to-text, open-source software, and the workflows Muesli is built to own.
An Otter.ai alternative for meeting notes without a bot
A Mac-first comparison for people who want meeting capture and local speech-to-text without inviting another participant into every call.
A Fireflies.ai alternative for local-first meeting notes
What it looks like when meeting transcription starts on the Mac already in the call instead of with a hosted meeting assistant.
TAKE THE SPEECH LAYER HOME
Read the source. Run the app. Keep the transcript close.
Muesli is open-source Mac dictation and meeting transcription built for local ASR on Apple Silicon.