Requirements

Requirements & Optional Helpers

TypeWhisper runs on a modern Linux desktop session. The core app is self-contained, but several features rely on optional desktop helper tools that you install through your distribution.

System requirements

• A modern Linux desktop session.
• The .NET 10 SDK — only needed to build from source. Prebuilt releases bundle the runtime; see Installation.
• ffmpeg — for file-transcription imports beyond the already-supported direct formats.

Optional desktop helpers

Each helper enables a specific feature. When one is missing, the Linux UI disables the related control and explains the reason, often with session-aware install hints.

Helper	Enables
pactl	Audio ducking
playerctl	Media pause during recording
pw-play (PipeWire) / paplay (PulseAudio) / aplay (ALSA)	Sound feedback
espeak-ng, espeak, or spd-say	Spoken feedback
xclip (X11)	Clipboard for clipboard-backed auto-paste
wl-copy / wl-paste (Wayland)	Clipboard for clipboard-backed auto-paste
wtype	Wayland keyboard input on wlroots compositors (Hyprland, Sway)
ydotool	Wayland keyboard input on GNOME / KDE Wayland (where wtype is unavailable)
xdotool	Keyboard input fallback on X11 and XWayland apps
busctl (systemd)	Toggling the session accessibility bridge for correction learning

Accessibility note: browser URL detection and Learn corrections from other apps both read text over AT-SPI, which needs the at-spi2-core bus (present by default on GNOME, KDE, and most desktops). Chromium/Electron and Qt apps additionally only expose an accessibility tree when the session flag org.a11y.Status.IsEnabled is on — the Dictation page offers a one-click Enable accessibility bridge button for that.

ydotool note: ydotool requires its daemon to be running and the current user to be in the input group. The Text insertion settings panel offers a one-click setup flow for the ydotool daemon and input-group membership when needed. See Wayland notes and Text insertion.

Optional CUDA backend

For GPU-accelerated transcription on compatible NVIDIA systems:

• An NVIDIA GPU and driver — the only hard requirement.
• The CUDA 12 runtime libraries (libcudart.so.12, libcublas.so.12, cuDNN, …) are downloaded on demand the first time you select NVIDIA CUDA and cached under ~/.local/share/TypeWhisper/Runtimes/ — only the libraries you are missing are fetched, and nothing GPU-related ships in the packages. You do not need to install a CUDA toolkit yourself.
• CUDA applies to the bundled whisper.cpp and sherpa-onnx (Parakeet, Canary) engines; other bundled local engines stay on CPU.

See GPU & CUDA for details.

Related pages

• Installation — release formats and build-from-source steps.
• Wayland notes — per-distro Wayland typing backend and global-hotkey setup.

---

Changelog

Date	Change
2026-06-17	Initial version.
2026-06-17	Corrected sound-feedback helper to pw-play / paplay / aplay.
2026-06-23	v0.10.0: CUDA runtime libraries are now downloaded on demand (GPU + driver is the only hard requirement); CUDA extended to the sherpa-onnx engine.
2026-07-27	v0.13.0: added the AT-SPI / accessibility-bridge note and busctl for correction learning.