Skip to main content

Voice Input settings

Settings → Voice Input controls how Kalam hears you and how transcribed text lands in other apps. Use it to pick a microphone, confirm permissions, tune silence detection and audio cleanup, and understand paste vs type insertion (including sensitive apps).

What you see

Three sections:

  1. Microphone — permissions, device picker, mic test
  2. Audio processing — silence detection and audio cleanup
  3. Text insertion — how text is delivered to the focused app

Microphone

Permissions & capabilities

A status panel shows whether Kalam can:

CapabilityWhy it matters
MicrophoneCapture audio for dictation
Accessibility (Mac) / App input (Windows)Insert text; on Mac, also read highlighted text for context-aware dictation
Global hotkeyRegister dictation shortcuts system-wide

When something needs action, use the buttons (for example Open Microphone settings, Open Accessibility settings, or Request accessibility prompt on Mac).

Device and test

SettingWhat it does
MicrophoneSelect your audio input device, or System default. Refresh the list if a new device was plugged in.
Test microphoneRecord a short sample, watch the level meter, then Play it back to check levels.
Mic not working?

Use Open microphone settings (or privacy settings on Windows), then Refresh device list. Speak while recording — the level bar should move.

Audio processing

SettingWhat it does
Silence detectionHow long Kalam waits for silence before ending the recording: Fast (0.8s), Balanced (1.5s), or Accurate (2.5s).
Audio cleanupPre-process audio before transcription (same chain as dictation and the mic test).

Audio cleanup presets:

PresetWhat it does
Off — no processingRaw microphone input
Light — gentle cleanup for most roomsBest default for most mics: trims low rumble, light compression, consistent volume
Moderate — stronger noise reductionBetter in noisier rooms; may sound more processed
Custom — tune each stepManual sliders: Bass roll-off, Background noise reduction, Volume leveling (ratio), Output loudness (normalize), Volume leveling (threshold)
Long or quiet speech

Long passages and quiet speech often benefit from Light or Moderate so levels stay even.

Text insertion

Dictated text is pasted into the focused app by default. Sensitive apps are typed instead so the transcript never hits the clipboard.

SettingWhat it does
DefaultHow text lands in most apps — Paste on Windows/Linux; on Mac, Accessibility, then paste.
Sensitive appsPassword managers and apps on your Privacy list are Type automatically. Manage the list under Settings → Privacy.

Clipboard behavior

Expand Clipboard behavior for details:

  • After paste, Kalam restores your previous clipboard (including images).
  • On Windows, dictation pastes are kept out of clipboard history and cloud sync.

See Sensitive apps and Context awareness.