Skip to content

Get started in five minutes

YazSes types what you say into whatever app has focus, and it does all of the speech recognition on your own computer. This page takes you from nothing to your first dictated sentence. Then it shows the three things most people try next.

What you need: Linux, macOS 11 or later, or Windows 10/11, with 4 GB of RAM (8 GB is comfortable) and a microphone. You do not need a GPU, an account or an API key. The platform support page lists every OS and CPU in detail.

1. Install

bash <(curl -fsSL https://raw.githubusercontent.com/MSKazemi/yazses/main/install.sh)

The universal installer works on Ubuntu, Debian, Fedora, Arch and their derivatives. It installs YazSes and runs yazses setup, which adds the audio, hotkey and text-injection helpers for your session (X11 or Wayland). It finishes by running yazses doctor.

Prefer a package manager? The Linux guide covers APT, Snap and pipx.

On Apple Silicon:

brew tap MSKazemi/yazses
brew trust MSKazemi/yazses
brew install --cask yazses

On an Intel Mac, download the x86_64 .dmg from the latest release. The macOS guide explains both routes and the first-launch Gatekeeper prompt.

winget install MSKazemi.YazSes

winget can trail the newest release by a few weeks. To be sure of the newest one, download the installer from the latest release, or use Scoop or Chocolatey (choco install yazses), which track every release. The build is not code-signed yet, so SmartScreen warns once. Choose More info → Run anyway. The Windows guide covers each route and updating.

pipx install yazses

This needs Python 3.11 or newer. On Linux, follow it with yazses setup, which installs the system helpers that a Python package cannot install itself.

2. Allow and check

yazses doctor

Every row should read [OK], and a row that does not names its fix. If doctor reports that you were added to the input group, log out and back in once. Group changes apply only to new logins.

Then tune the voice gate to your microphone and room:

yazses mic-level --set

Open YazSes from Applications. The app is not notarised yet, so the first launch is blocked: on macOS 15 and later, approve it with Open Anyway in System Settings → Privacy & Security; on macOS 14 and earlier, right-click the app and choose Open.

macOS gates the three things a dictation tool needs behind three separate switches. All three are in System Settings → Privacy & Security:

  1. Accessibility, so YazSes can type into other apps.
  2. Input Monitoring, so it can see the hotkey. This is the one people miss. With only Accessibility on, the key does nothing.
  3. Microphone, so it can hear you.

Quit and relaunch YazSes after switching them on. See why there are three switches.

YazSes starts in the system tray, next to the clock. If Windows asks whether YazSes may use the microphone, allow it. The tray icon's colour shows its state at a glance: see what the colours mean.

yazses doctor

It checks the microphone, the hotkey, the text-injection backend and permissions, and names the fix for anything that is missing.

3. Dictate

On macOS and Windows the app you opened is already running, so skip to the hotkey. On Linux, or after a pipx install, start the daemon:

yazses start

Then hold the hotkey, speak, and release:

OS Default hotkey
Linux Right Alt
macOS Right Option
Windows Right Ctrl

The text appears in the focused app one to two seconds after you let go (a median of 1.6 s on a laptop CPU with the default model). The hotkey is a modifier you rarely use, so it does not interfere with normal typing. You can also choose a different key.

The first dictation downloads the speech model

The first time YazSes transcribes, it downloads the Whisper model once. After that, dictation never uses the network. To set up a machine that is never online, see running fully air-gapped.

What to try next

  • Voice commands: say "undo that", "save file" or "go to line 42" and YazSes sends the keys instead of typing the words.
  • Transcribe a recording: yazses transcribe meeting.m4a produces text or subtitles offline, with optional speaker labels.
  • Teach it your words: names, jargon and identifiers it keeps mishearing.
  • Pick a model: trade a little speed for accuracy, or the reverse on a low-RAM machine.
  • The Settings window: every feature is a checkbox, so you do not need to edit a config file.
  • Troubleshooting: start with yazses doctor, then find your symptom.

If something does not work

Run yazses doctor first. It covers most first-run problems, and every check that fails names its fix. The common ones:

  • Nothing happens when I hold the key. On Linux, log out and back in after the installer adds you to the input group. On macOS, turn on Input Monitoring, then relaunch.
  • It types into the wrong window, or nowhere. See dictation types into the wrong window.
  • "Silent audio — discarding". The voice gate is set above your voice. Run yazses mic-level --set again; see the guide.

Still stuck? Open an issue and paste what yazses doctor printed. Read it first: it lists device names and file paths.