Using SigmaShake Voice
Basic dictation
- Press your hotkey — the voice indicator (pill) appears in the corner of your screen. You're now recording.
- Speak at a normal pace.
- Press the hotkey again — the app transcribes your speech and pastes the text wherever your cursor is.
The whole cycle takes about 1–3 seconds depending on how long you spoke.
Settings
Open Settings by right-clicking the tray icon → Settings.
Hotkey
Click the Hotkey field to enter recording mode (the button pulses blue), then press the key you want to use.
Recommended keys: F9, F10, Pause, ScrollLock
Press Esc to cancel without saving. Click Clear to remove the hotkey entirely.
Use a single bare key like F9 — not modifier combos. See Install → Wayland for details.
Input Device
Select the microphone to use. The System default option follows your OS audio settings. If you have a dedicated mic (USB headset, audio interface, etc.), select it by name.
Sound effects
Enable or disable the startup/stop audio chirps. Toggle in Settings → Behavior. Changes take effect immediately — no restart needed.
Notification verbosity
Controls which events send system notifications:
| Level | Notifications sent |
|---|---|
| Quiet | Errors and transcription results only |
| Normal (default) | Listening, Transcribing, Done, Errors |
| Verbose | All of the above plus timing details |
Milestone toasts (word-count badges, streak records) always fire regardless of this setting.
Auto-paste
When enabled (default), text is automatically pasted at your cursor after transcription via a simulated Ctrl+V. When disabled, the transcript is placed in your clipboard and you paste manually.
Custom dictionary
Add words that Whisper frequently misspells — brand names, technical terms, names. One entry per line, max 50. The words are injected as a vocabulary hint at the start of each transcription.
Example: adding kubectl, sigmashake, WebSocket means those words come through correctly even if Whisper's default would mangle them.
Punctuation voice commands
When enabled, spoken commands are replaced with their punctuation equivalents:
| Say | Gets inserted |
|---|---|
period | . |
comma | , |
question mark | ? |
exclamation point | ! |
new line | newline |
new paragraph | blank line |
open quote / close quote | " |
open paren / close paren | ( / ) |
dash / hyphen | — / - |
colon / semicolon | : / ; |
Escaping: prefix any spoken command with literal to pass it through unchanged: "the Tudor literal period" → the Tudor period.
Auto-capitalisation fires after ., !, ?, and newlines.
Pill color theme
Choose the accent color for the recording waveform from five presets:
- Red (default), Blue, Green, Purple, Mono (greyscale — useful for screen recordings)
The idle dim-grey state is unchanged across themes. Transcribing always uses blue; errors always use amber.
Wake word
Set a word or phrase to start recording hands-free. When configured, the app listens continuously and begins recording automatically when it hears the wake word.
- Uses CPU in the background while listening
- Leave blank to disable (default)
- Example: "computer" or "hey sigma"
End word
Set a word or phrase to stop recording and trigger transcription. The end word is stripped from the final transcript.
- Leave blank to disable (default)
- Example: "done" or "stop"
Stats and streaks
Open tray → Stats… to see your dictation dashboard:
- Total words and total sessions dictated
- Time saved — estimated minutes recovered vs. typing at 200 WPM
- Current streak and longest streak (consecutive days with at least one dictation)
- Milestone badges — unlock for 100 / 1 k / 10 k / 100 k words and other milestones; each fires a one-time notification
The Reset stats button in the dashboard (double-click to confirm) clears all counters.
Stats can be disabled entirely in Settings → Privacy & Data.
Recording history
After each dictation the transcript is saved to a history log (newest first, up to 50 entries). View it in Settings → Privacy & Data. Click Clear history to wipe it.
History can be disabled in Settings → Privacy & Data.
The voice indicator (HUD pill)
A small floating panel appears in the corner of your screen while recording. It shows:
- Waveform bars — real-time mic level visualization
- State — Listening → Transcribing → done (pill disappears)
Moving the indicator — click and drag the pill to any position. It saves its position automatically and reappears there next time.
Snap-to-corner — when enabled (default), the pill snaps to the nearest screen corner when you release it within 80 px of one. Turn off in Settings → Behavior for free positioning.
On Wayland compositors, the window may appear at a compositor-chosen position the first time. Drag it to your preferred spot and it will stay there.
Always-on indicator
Enable Always show indicator pill in Settings (tray → Settings → Behavior) to keep the pill visible at all times — not just during recording.
When always-on is active:
- The pill shows a quiet dim dot when idle and switches to the animated waveform when recording starts.
- Repositioning — hover over the pill (on Linux X11, macOS, Windows) and drag it to a new spot. It becomes interactive when your cursor is near, and click-through when your cursor is away so it doesn't block anything underneath.
- The saved position is shared between idle and recording — wherever you drop it is where it will appear.
On Wayland, the pill cannot detect the cursor position and pass clicks through. When always-on is enabled it stays interactive, so park it in a low-traffic corner (the default bottom-right position works well).
Notifications
SigmaShake Voice sends a system notification at each stage:
| Event | Notification |
|---|---|
| Start recording | "Listening… press hotkey again to stop" |
| Transcribing | "Transcribing…" |
| Done (auto-paste) | Text preview with "✅ Pasted" |
| Done (clipboard only) | Text preview with "📋 Copied — press Ctrl+V" |
| Error | Description of what went wrong |
Startup notifications will alert you if the microphone is unavailable or the Whisper model failed to load, so you don't need to check logs.
First-run tutorial
On first launch a setup wizard walks you through:
- Hotkey — press any key to set it
- Microphone — pick your input device; a live meter confirms it's working
- Test dictation — press the hotkey, say something, see your words appear
- Pill position — drag the indicator to your preferred corner
Skip any step or click Skip all to dismiss. Reopen the wizard at any time via tray → Help — Tutorial.
Startup behavior
The app starts with the tray icon only — no window. It pre-arms the audio stream and loads the Whisper model in the background (takes 1–3 seconds on first run). The tray icon is ready immediately; if you press the hotkey before the model loads, you'll see a "Whisper model loading" notification and can try again in a moment.
Logs and config location
| Platform | Location |
|---|---|
| Linux | ~/.config/sigmashake-voice/ |
| macOS | ~/Library/Application Support/sigmashake-voice/ |
| Windows | %APPDATA%\sigmashake-voice\ |
The Settings page footer shows the exact path for your system.
Log files are named voice.YYYY-MM-DD.log and rotate daily.