The WinForms .NET 10 client uses VoiceCat.Core for TLS, TOFU, protocol state and encrypted UDP, and VoiceCat.Audio for streams, jitter, Opus and mixing. Its only native runtime component is voicecat_media.dll, the narrow Opus/RNNoise C shim. The old voicecat.dll core is no longer loaded or published.

Build

Stage the pinned native media dependencies once, then build the solution:

./dotnet/build-native.ps1
dotnet restore clients/windows/VoiceCat.slnx --locked-mode
dotnet build clients/windows/VoiceCat.slnx -c Release --no-restore

The original VoiceCat.Interop project remains in the repository as a migration oracle. The app references VoiceCat.Managed, whose compatibility facade lets the existing accessible WinForms UI keep its event-pump shape while all networking and audio state live in the idiomatic managed libraries.

Publish

./clients/windows/publish-client.ps1

This produces a self-contained win-x64 distribution in dotnet/artifacts/client/win-x64. The script requires voicecat_media.dll and fails if the legacy voicecat.dll appears.

The noninteractive startup and real WASAPI device check is:

./dotnet/artifacts/client/win-x64/VoiceCat.App.exe --smoke-test --audio

--smoke-test constructs the real main form and pumps the managed client. --audio additionally opens the default WASAPI capture and render endpoints, moves PCM through both for three seconds, and fails if capture produces no samples.

Run manually

# Terminal 1
./dotnet/artifacts/server/win-x64/VoiceCat.Server.exe

# Terminal 2
dotnet run --project clients/windows/VoiceCat.App/VoiceCat.App.csproj

Manual release validation includes NVDA navigation and a sustained two-client listen test, as tracked in docs/roadmap.md.