Skip to main content
Kalam Documentation

Voice Input settings

Settings → Voice Input controls how Kalam captures your microphone and prepares audio before transcription, and shows how text is inserted into focused apps on your operating system.

Settings → Voice Input showing microphone checklist, device picker set to System default, Test microphone controls, and level meter. (Click to zoom)

An attention dot on the Voice Input tab indicates a missing microphone permission or unavailable input device.

Microphone & permissions​

PlatformChecklist behavior
WindowsDone confirms Kalam can open your microphone cleanly.
macOSLists Microphone, Accessibility, and Input Monitoring grants with direct OS settings shortcuts (or repair prompts after an app update).
LinuxLists detected audio devices directly (Found means a usable input is available).

Once a microphone path is ready:

  • Pick System default or a specific headset/USB mic from Microphone (click Refresh after plugging in a device).
  • Click Record under Test microphone, speak a test sentence, and click Play to verify input gain and audio cleanup before recording real notes.

Audio processing​

SettingOptionsWhat it does
Silence detectionFast (0.8s), Balanced (1.5s), Accurate (2.5s)Wait time after speech stops before automatically ending a toggle or silence-trimmed capture. Default Balanced.
Audio cleanupOff, Light, Moderate, CustomFilters low-frequency rumble and evens out speaking volume before recognition runs. Default Light.

Select Custom to tune Bass roll-off (40–200 Hz), Background noise reduction (dB), Volume leveling (ratio / threshold), and Output loudness (normalize).

Text insertion status​

This card reports how Kalam delivers text on your operating system:

TargetmacOSWindows & Linux
Normal appsAccessibility, then paste — tries native Accessibility insert first and falls back to clipboard paste.Paste — places the transcript via the OS clipboard (Wayland COSMIC uses native IME when supported; GNOME/KDE Wayland prompt once for remote interaction).
Sensitive appsType automatically — simulates keystrokes so transcripts never touch the clipboard.Type automatically — simulates keystrokes so transcripts never touch clipboard history or cloud sync.

When Kalam uses clipboard paste, it restores your previous clipboard contents (including images) immediately afterward.