StarWhisper

Windows Sandbox screen recording / 42 seconds

Watch StarWhisper process a recording offline.

This app-only capture follows one StarWhisper run from the listening state to a completed sentence. The Windows Sandbox reported networking disabled throughout the visible run. The scope is deliberately narrow: it documents this build and this test condition, not a universal promise about every computer.

Silent app-only screen recording. Captions describe visible labels, live bubble text, and the final inserted sentence.

What this proves, and what it does not

The captured app state and the adjacent run record agree on the test condition. The page does not stretch that evidence into a claim about all hardware, models, input paths, or future builds.

  1. 1. Local modeThe app labels the captured build as LOCAL MODE.
  2. 2. Network conditionThe app shows the Windows Sandbox disabled-network label, and the run record reports no network reachability.
  3. 3. Text placementThe live bubble completes, then the sentence appears under Dictation while the labels stay on screen.

Visible result text

“Press one key, speak normally, and StarWhisper puts the complete sentence right into the app I am using.”
This is a visual transcription of the result shown in the recording. There is no audio stream in this 42-second export.

Before you try it

Prerequisites

  • A 64-bit Windows 10 or Windows 11 computer.
  • StarWhisper installed with a local speech model available.
  • A microphone and permission for a real live test on your own machine.

Limits of this evidence

  • The clip shows the complete captured run, but it does not show the Windows network setting being toggled by hand.
  • The run used a pre-recorded audio fixture, so it does not prove live microphone accuracy.
  • The capture shows one build, one model path, and one sentence.
  • The footage is historical. Check the current download page for current setup and plan terms; this page changes neither pricing nor entitlements.

The target here is a demo text field, not Notepad. The input was a pre-recorded local audio fixture rather than a physical microphone. For a current setup walkthrough, see the offline dictation guide. The test lab provides separate licensed audio and raw engine results.

Want to run the local-mode check yourself?

Download StarWhisper