Imagine your music is playing and you want to talk. You say, "A little quieter, please!" The music turns down, lets you finish, then comes back. That is the whole idea. No shouting match with the speakers.

I talk to AI while Spotify keeps me company. Windows' Communications setting did not lower Spotify the way I needed, and reaching for the volume slider every time got old. So I made a tiny volume referee: VoiceDucker.

I press Enable when I want it to listen to my microphone. When I speak, it lowers Spotify; I can choose other playing apps too. When I stop, it fades them back to the volume I chose. My voice stays in memory: no recordings, transcription, or uploads.

VoiceDucker is a native Windows desktop app built with WinUI 3, the same UI framework we use for our newer Windows apps. We pay close attention to RAM usage, performance, and disk footprint. There is always room to improve, and we keep working on optimizations.

Created by Umberto Giacobbi, shared under the MIT license.

The first public version is 0.1.0, tagged v0.1.0-stable. Get the matching ZIP from GitHub Releases:

• Windows x64: VoiceDucker-win-x64.zip, for most Intel and AMD PCs.
• Windows ARM64: VoiceDucker-win-arm64.zip, for Windows on ARM.

On Windows 11, extract the whole ZIP into a folder you can write to, then open VoiceDucker.exe. Keep its companion files together; the ZIP includes the .NET and Windows App SDK runtimes. The microphone starts off. Put some music on, press Enable, and talk at your usual volume. Adjust the threshold if it needs a nudge.

The ZIP is portable and unsigned. Windows may show a security prompt. There is no MSIX installation or signing certificate to import for this download.

New UI and settings · Open UI screenshot · Open settings screenshot

• Reduce volume by 90%
• Fade down in 300 ms
• Fade back to my previous volume in 30,000 ms
• Start with Windows, with the microphone off until I press Enable
• Lower Spotify by default; other apps are optional

Every valid audio setting is saved immediately, including while I type a number, and restored after app or Windows restarts. Windows startup changes take effect as soon as I toggle the checkbox. The microphone still starts off.

Each app appears as a simple row without a card or Remove button. Its live session peak meter overlays the volume slider, with a -60 to 0 dBFS display range. The first Advanced option is Sensitivity, a full-width slider from -24 to +36 dB. It applies software gain after microphone capture, before sound detection and the microphone meter. It updates immediately and is saved across restarts. It does not change the Windows microphone volume or the input used by other apps. Any sound above the resulting detection threshold can lower playback.

Reset next to Sensitivity returns microphone gain to 0 dB. The microphone level meter uses a thick green bar, shows the live post-gain signal from -60 to 0 dBFS, and marks the detection threshold at about -35 dBFS, so adjusting sensitivity has a visible reference. The meter labels whether the current sound is above or below the threshold. While the microphone is off, it prompts me to press Enable to see the level. The Sensitivity slider uses the same green color.

Microphone lets me select an active input or use the Windows communications default; the main window shows the actual device in use. Selecting another input while enabled restarts capture on that device. If a saved device is unavailable, VoiceDucker reports it rather than silently switching microphones. The list refreshes when I open the selector. Selecting an input while disabled does not start capture.

Wait before fade back (ms) keeps audio lowered during short pauses, starting from the last detected sound. The default is 650 ms, adjustable from 0 to 10,000 ms. After that wait, the existing Fade back duration controls the return to my volume. Changing the wait takes effect immediately. Both the microphone choice and wait are saved across restarts.

I can change these settings in the window. If I move an app's volume slider myself, VoiceDucker leaves my new choice alone. On a normal exit, it returns the sessions it still controls to their previous levels over 2,000 ms.

VoiceDucker's own audio sessions are filtered by process name. Existing hidden apps remain excluded; Hidden apps in Advanced options lets me restore them.

See release and WinGet instructions for packaging and publication.

On Windows 11 with the .NET 10 SDK and Windows App SDK build tools, I run:

.\scripts\build.ps1 -Architecture x64 -Configuration Debug -Run

The build puts the app at artifacts/app/win-x64/VoiceDucker.exe. Build ARM64 with -Architecture ARM64. I publish only portable x64 and ARM64 ZIPs through GitHub Actions, never MSIX packages there.

For a local signed Release MSIX, I use a trusted code-signing certificate with subject CN=VibeWare in Cert:\CurrentUser\My:

pwsh -NoProfile -File .\scripts\package-msix.ps1 -Architecture x64 -CertificateThumbprint <thumbprint> -Install

The installed MSIX keeps the Start menu identity VibeWare.VoiceDucker_af9ft7172qtwj!App across updates and registers the VoiceDucker.exe execution alias. Use the installed Start menu entry or this alias to launch the installed app. The portable build uses the stable path artifacts/app/win-x64/VoiceDucker.exe; temporary validation builds stay under artifacts and do not create launch shortcuts.

VibeWare is a term coined by Umberto Giacobbi and an open initiative for developers who build with AI and care about the craft. It challenges the assumption that vibe coding is synonymous with low-quality code. It openly acknowledges the weaknesses and risks of AI-generated software: experienced developers must guide the process, test carefully, check security and take responsibility for the result. A VibeWare footer simply says: this software was developed with AI, and it deserves to be judged by the quality of the work.

Read the VibeWare manifesto

You are welcome to explore, adapt and reuse the open-source VibeWare materials under the project license — my contribution to the developer community.