SocketJack.Unity.Voice.Web.Client 2026.11.4

SocketJack.Unity.Voice.Web.Client adds account-backed audio over a separate authenticated /ws/voice connection. It depends on SocketJack.Unity.Voice and SocketJack.Unity.Web.Client, and contains no server or SQLite implementation. The server integration is a separate SocketJack.Unity.Voice.Web package.

For Unity, enable the optional Web Client, Voice and Voice Web Client modules in the package bundle. Add SocketJackWebClientBehaviour, SocketJackVoiceManager, SocketJackWebVoiceClient and optionally SocketJackVoiceControls to a scene UI object. The voice client follows account sign-in and connects to the same server origin. It uses the account session only in the first bounded text frame; tokens are never placed in URLs or serialized scene fields. Native clients use the platform TLS certificate validation. WebGL uses the included SocketJackBrowserVoiceSocket component for text-authentication/binary-audio WebSockets, plus SocketJackBrowserMicrophone for Web Audio capture/playback. Native clients keep normal platform certificate validation. Browser capture/audio requires explicit user permission and may need a click/tap to resume audio.

Microphone capture remains off until StartCapture() or the player's Enable microphone button. StopCapture() stops it. Push-to-talk defaults to V with the legacy Input Manager; new Input System actions call SetPushToTalk(bool). Channel selects Proximity, Party, Raid, Community, Whisper or Global. ScopeId selects the group/community, and TargetAccountId selects a whisper recipient. The optional chat panel supplies selected group/community and sender IDs. Choosing a channel never grants membership.

Without Unity, create AccountVoiceClient from a signed-in SocketJackWebClient, await ConnectAsync, then submit bounded 20 ms Opus frames with SendEncoded. FrameReceived provides canonical account identity, channel and audio. Feed audio through VoiceJitterBuffer/VoicePcmBuffer for playback. Custom capture adapters use the Voice module's codec. An optional SocketConnector supports a platform WebSocket implementation; it must preserve normal certificate validation for remote WSS endpoints.

Audio queues retain at most four outgoing frames. Stream sequences reject duplicates/reordering at playback, and server-clock metadata discards packets older than 500 ms rather than playing a stale TCP/WSS backlog. Socket congestion can still cause reliable-stream head-of-line blocking. Use native game-session unreliable UDP voice where appropriate for lower latency. No audio is written to chat history or account persistence.

NuGet provides portable DLLs and component source as content; it does not automatically import a Unity project. For Unity, use the optional package bundle or copy its DLLs/components as described in the main setup guide. Standalone .NET projects can use the NuGet dependency graph directly.