Entune View on GitHub

Entune guide · Offline

Dictation without an internet connection.

Once Entune and a local speech model are installed, Entune is built to turn your voice into text on your computer with no network connection. Here is how to set it up, and what still needs the internet.

Free and open source · Local speech models · Mac, Windows and Linux

Updated

Set it up See what it does →

Set it up while you’re online

  1. Install Entune with the command below.
  2. Download and select a local speech model in Models: a Whisper.cpp model on macOS, Windows or Linux, or Parakeet on an Apple Silicon Mac after its one-time engine install (uv tool install parakeet-mlx). Downloading a model does not always make it the one Entune uses, so confirm it is the selected speech model.
  3. Allow the permissions Entune asks for, and set a shortcut.
  4. Keep the optional processing off: corrections, filler removal and formatting, in Settings › Corrections & formatting. They start off.
uv tool install entune && entune

Needs uv, which installs Python for you, or pip with Python 3.12+. Free and open source.

Puts Entune in Applications and opens it. The first start can take up to a minute.

What works without a connection

With a local speech model and the optional processing off, Entune is built to dictate with no network connection:

  • Starting Entune: it checks for no updates and its window loads only Entune’s own files, with no web fonts or online scripts.
  • Transcribing: the model you downloaded loads from Entune’s folder on your computer.
  • Dictating: recording, transcribing and pasting make no network call.

That is how Entune is designed; a complete dictation session with the network switched off hasn’t been tested on record yet. Before you rely on it, try a short dictation with Wi-Fi off.

What still needs the internet

  • Downloads: installing Entune, the speech models, and the engines for Parakeet and Laya.
  • Cloud speech models: AssemblyAI, Groq, Soniox, ElevenLabs and xAI run on their providers’ servers.
  • Cloud decision models: Jev, OpenAI’s Decisions API and Perplexity, when you turn on corrections, filler removal or formatting.
  • Building your dictionary: it always uses a cloud language model, and runs only when you ask.
  • Laya: the decision model that runs on your computer downloads its model the first time you use it. Whether it then starts with no connection at all hasn’t been tested.

How this compares with built-in dictation

On a Mac, Dictation’s settings show whether text dictation is processed on your Mac. Windows voice typing needs an internet connection, because it uses Microsoft’s online speech recognition. Most Linux desktops include no system-wide dictation. What is dictation software? covers each in more detail.

For choosing a local model and keeping the optional processing separate, read set up local dictation. On Linux, start with dictation on Linux.