Coming soon
Dictate and read text aloud, on Linux, with your own API key.
Press a hotkey in any application, speak, and the text appears where your cursor is. Press another to have selected text read back to you. keyspeake is being built now — join the waitlist to hear the moment it's ready.
Bring your own API key — Groq, Deepgram, or OpenAI. No keyspeake account, no middleman. No spam, one email when it launches.
Dictation
Hold a hotkey, speak, and the words land as real keystrokes in whatever has focus — a terminal, an editor, a browser field, a chat window. keyspeake types through a virtual keyboard, so it respects your keyboard layout instead of assuming US QWERTY, and it works in places clipboard-paste tricks don't, including terminal apps running raw-mode TUIs.
Read-aloud
Select any text and have it read back to you — same hotkey pattern, same one keybind. Playback shows up as a small on-screen player with pause, seek, speed control, and hardware media-key support, so it behaves like a real audio player, not a one-shot text-to-speech demo.
Built for how Linux actually works
Not a generic wrapper that happens to launch under Wayland. keyspeake talks to your window manager directly:
Hyprland
Native IPC for window tracking and terminal-aware paste.
Sway
swayipc for the same window-focus awareness.
GNOME
D-Bus integration for window tracking under Mutter.
KDE
D-Bus integration for window tracking under KWin.
X11
Direct X11 window tracking, no compositor dependency.
Text injection goes through a virtual uinput keyboard with real XKB layout resolution — not a clipboard hack — so dictation works correctly on non-US layouts, and in terminal emulators where clipboard shortcuts send control characters instead of pasting.
Your key, your account, no middleman
keyspeake doesn't run a server. You configure your own API key for the transcription and text-to-speech provider you choose — audio goes directly from your machine to that provider. There's no keyspeake backend in between deciding what your dictation is worth or storing a copy of it.
- Crash reporting is opt-in and off by default — nothing is collected unless you turn it on, and it stays local.
- There is no telemetry or usage analytics anywhere in the product.
- Dictation history and cached audio live on your disk. Saving audio at all is a setting, not a default.
What it costs to run
keyspeake itself has no announced price yet. The only cost while it's in development is your own API usage, billed directly by whichever provider you pick:
| Backend | Pricing model | Typical cost for casual daily use |
|---|---|---|
| Groq | Free tier, then usage-based | Free tier covers most single-user dictation |
| Deepgram | $200 free credit on signup, then usage-based | Free credit lasts months of regular use |
| OpenAI | Usage-based, no free tier | Depends on model chosen — realtime streaming costs more than batch |
| Local whisper.cpp | Free — runs on your CPU/GPU | $0, offline, slower on modest hardware |
Exact cost depends on how much you dictate and which model you pick — these are the providers' own published terms, not a keyspeake fee.
What this doesn't do yet
- Linux only. No macOS or Windows build. Not currently planned.
- You bring your own API key. There's no managed/bundled key option — one more account to set up before you can dictate.
- No mobile companion. Desktop only.
- Vosk and Parakeet are early. whisper.cpp is the accurate offline option today; Vosk trades accuracy for a much smaller (~40MB) model, and Parakeet (NVIDIA GPU) isn't finished.
FAQ
Do I need an API key?
For cloud backends (Groq, Deepgram, OpenAI), yes — you sign up with the provider and paste your own key into keyspeake's setup wizard. If you'd rather not, the local whisper.cpp backend runs fully offline with no key and no account.
Does it work offline?
Yes, using the bundled local whisper.cpp engine, or the smaller Vosk model if you'd rather trade some accuracy for a lighter footprint. No internet connection or API key required for either.
Which distros are supported?
Any modern Linux distro on x86_64 or aarch64 — Arch, Debian/Ubuntu, Fedora, NixOS, and anything you can build from source on. See the download page for install methods per distro.
When is keyspeake coming out?
No fixed date yet — it's actively being built. Join the waitlist above and you'll get one email the moment it's ready to download.
Will macOS or Windows come?
Not currently planned. keyspeake's integration depth (uinput, XKB, per-compositor window tracking) is specifically a Linux architecture, not a cross-platform abstraction with Linux support bolted on.