System requirements

Kudu is light: a small app plus a speech model it downloads once. Here’s what each platform needs.

macOS

ChipApple Silicon: M1, M2, M3, M4 or newer. Intel Macs aren't supported.
SystemmacOS 14 Sonoma or newer
Smart formatting & Command modemacOS 26 with Apple Intelligence turned on
Memory8 GB (16 GB recommended)
Disk~15 MB app + ~450 MB speech model, downloaded once
PermissionsMicrophone and Accessibility

Windows (Beta)

SystemWindows 10 (22H2) or Windows 11, 64-bit (x64)
Processor4 cores with AVX2 (most CPUs from 2015 on)
Memory8 GB (16 GB for the Better smart-formatting model)
Disk~15 MB app + 250–480 MB speech model; smart formatting 0.5–1.1 GB, optional
Not yetWindows on ARM

Linux (Beta)

Processorx86_64 with AVX2, 4 cores
Systemglibc 2.35 or newer (Ubuntu 22.04, Debian 12, Fedora 38, Arch…)
DesktopX11 or Wayland. The tray needs StatusNotifierItem (KDE; GNOME with the AppIndicator extension).
AudioPipeWire, PulseAudio or ALSA
Memory8 GB (16 GB for the Better smart-formatting model)
AccessYour user in the input group, and /dev/uinput (the installer sets this up)

Downloads

WhatSizeWhen
Kudu for Mac~6 MB download, ~15 MB installedInstall
Kudu for Windows / Linux15–22 MBInstall
Speech model on a Mac (Parakeet)~450 MBFirst launch
Speech model on Windows / Linux: Accurate (Parakeet)482 MB download, 631 MB on diskFirst launch
Speech model on Windows / Linux: Compact (Moonshine)251 MB download, 272 MB on diskFirst launch
Smart formatting on Windows / Linux: Small (Qwen2.5 0.5B)491 MBOptional
Smart formatting on Windows / Linux: Better (Qwen2.5 1.5B)1.1 GBOptional

Which speech model should I pick on Windows or Linux?

Accurate unless your PC is older or short on memory. Both run fully offline; Accurate makes fewer mistakes with names and fast speech, Compact is smaller and quicker. You can switch any time in Settings.

Ready? Download Kudu or follow the install guide for macOS, Windows or Linux.