Read aloud
Select text in any app and press ⌃⌥S. FreeFlow reads it back in Kokoro's voice, generated on your Mac with nothing sent anywhere. Nothing selected? It reads your clipboard. Press ⌃⌥S again, or Esc, to stop. The voice is a one-click 117 MB download under Models, English only, and the reading speed lives in Settings → Read aloud.
Why a dictation app reads
Hearing a draft is the fastest way to find the sentence that does not work, and the people who dictate most are often the same people who would rather listen than re-read. macOS has a built-in speak-selection feature; Kokoro sounds considerably more like a person, and FreeFlow puts it one keystroke away in every app.
- Download the voice onceModels → Kokoro → Download. 117 MB, cached by FluidAudio in
~/.cache/fluidaudio/Models/so other apps built on the same library can share it. Delete it from the same place. - Select some textIn Mail, a browser, a PDF, Notes — anywhere a selection is possible. Nothing selected means the clipboard is read instead.
- Press ⌃⌥SOr use Read Aloud from the menu bar. FreeFlow copies the selection with a synthesized ⌘C (that is why it needs Accessibility), then speaks it. Your clipboard is put back afterwards.
- Stop with ⌃⌥S or EscReading stops at once. Start again and it reads from the beginning of the current selection.
Settings
- Reading speed — Settings → Read aloud. 1× is the voice's natural pace; most people settle between 1.1× and 1.4× for proofreading.
- The shortcut — Settings → Hotkeys → Read selection aloud. FreeFlow warns if the key you pick is already claimed by another app.
- Slow apps — since 1.7.0 FreeFlow waits longer for an app to hand over the selection, which fixed read-aloud starting with stale clipboard text in heavy Electron apps.
Limits
- English only. Kokoro is an English voice. Other languages are read with an English accent, or not well at all.
- Not a screen reader. It reads what you select; it does not navigate or describe the screen. VoiceOver is the tool for that.
- Long selections take a moment to start while the voice is generated; it streams as it goes.
Questions people ask
Is the voice generated on my Mac?
Yes. Kokoro runs through FluidAudio as a CoreML model, the same way the speech models do. Nothing is sent anywhere.
Can it read a whole web page?
Select the page's text (⌘A in the content area) and press ⌃⌥S. Very long selections start after a short pause.
Why does it need the Accessibility permission?
To copy your selection with a synthesized ⌘C. FreeFlow restores the clipboard afterwards.
Can I choose a different voice?
Not yet. Kokoro is the one voice shipped; it is small, fast and sounds natural in English.