Linux ships as inkspoke.AppImage — a single self-contained executable. There is no .deb, no .rpm, and no Flatpak on the release page.
!
Set expectations before you start: Linux is the least-exercised of the three platforms. There is no GPU acceleration for local transcription — it runs on CPU only, so larger models are slow. Prefer a smaller model or a cloud one.
Running it
terminalchmod +x inkspoke.AppImage ./inkspoke.AppImage
Tested against Ubuntu 22.04 and Fedora 38 or equivalent. You need either PulseAudio or PipeWire running.
Audio and input access
Linux has no per-application permission prompts. Instead, access is decided by group membership, and InkSpoke can only tell you what is missing in its logs. If the microphone is not detected, or global hotkeys never fire, add yourself to the relevant groups:
terminalsudo usermod -aG audio $USER # microphone capture sudo usermod -aG input $USER # global hotkeys via /dev/uinput
✺
Group changes only apply to new sessions — log out and back in, then retest. This catches almost everyone at least once.
ffmpeg
Importing audio files and compressing audio for cloud transcription both need ffmpeg. The Linux build falls back to a system ffmpeg on your PATH, so install it from your package manager if those features misbehave.
Where your data lives
linux~/.config/InkSpoke/