Azure/azure-sdk-for-net Azure.AI.VoiceLive

1.2.0 (2026-08-03)

Added AzureRealtimeNativeVoice and AzureRealtimeNativeVoiceName with 12 new voice options (Aarti, Andrew, Ava, Denise, Diya, Elsa, Florian, Francisca, Meera, Xiaoxiao, Ximena, Yunxi).
Added AllowParallelToolCalls property on VoiceLiveSessionOptions.
Added Channels and ReferenceSource (EchoCancellationReferenceSource) properties on AudioEchoCancellation.
Added streaming text input events: ClientEventInputTextDelta, ClientEventInputTextDone.
Added ServerEventResponseInvocationDelta for streaming invocation deltas.
Added ExpiresOn property on VoiceLiveSessionResponse (server-set session expiration time).
Added the 2026-07-15 GA service version (VoiceLiveClientOptions.ServiceVersion.V2026_07_15), which is now the default.

Removed the following preview-only types that were added in 1.2.0-beta.1 and never shipped in a stable release: SmartEndOfTurnDetection (end-of-utterance detection with ThresholdLevel and TimeoutMs); RTC call support (ClientEventRtcCallSdpCreate, ServerEventRtcCallSdpCreated, ServerEventRtcCallError, RtcCallErrorDetails); and output audio buffer lifecycle events (ServerEventOutputAudioBufferStarted, ServerEventOutputAudioBufferStopped).
Updated samples to use MCPApprovalKind.