Skip to main content
Kalam Documentation

Privacy settings

Settings → Privacy controls how long dictation history and local audio recordings stay on this computer, which apps enforce fail-closed on-device protection, and anonymous usage telemetry settings.

Settings → Privacy showing History Retention set to 30 days, Keep recordings on, and Sensitive app detection on. (Click to zoom)

History & local recordings​

SettingDefaultOptions & behavior
History Retention30 daysChoose 7 days, 30 days, 90 days, 1 year, or Forever. History stays local on this device and never syncs.
Keep recordingsOnSaves audio locally so you can replay recordings in History with word-synced highlighting. Audio files are never synced or uploaded for storage. Switch off to discard audio immediately after transcription.
Keep audio for7 daysChoose 1 day, 7 days, 30 days, or 90 days (cannot outlive History Retention unless history is Forever).
Recording qualityStandard (Opus)Standard (Opus) (~11 MB per hour of speech) or Lossless (FLAC) (~50 MB per hour).

Sensitive apps​

  • Sensitive app detection (Default On) matches protected process names—including major password managers out of the box—plus any custom apps you add with Add app.
  • Matched windows force On-device speech recognition (failing closed if no local model is installed), block Read surroundings, block Commands, and type text directly via keystrokes instead of touching the clipboard.
  • See Sensitive apps for step-by-step setup and details on how Improve behaves in sensitive windows.

Telemetry​

Build channelDefaultBehavior
StableOffOpt-in anonymous usage metrics only (no audio, no transcripts, no API keys).
Beta (pre-release)On (locked)Automatically sends anonymous usage events and sanitized diagnostic logs under the Beta Policy to help catch regressions before stable release.
Switching channel on a beta binary

Changing Channel to Stable in Settings → About changes where updates are checked, not the binary currently running—beta telemetry stays active until you upgrade onto an actual stable release. See Updating and the beta channel.

How Kalam handles voice data​

Voice audio is processed in memory and—when using cloud speech engines—streamed directly to your selected provider. Kalam's own servers do not store your transcripts or local audio recordings. For full legal commitments, see the Privacy Policy.