Events & Delegates¶

All Aboleth STT events are declared as DECLARE_DYNAMIC_MULTICAST_DELEGATE and exposed as BlueprintAssignable properties. They are available on the subsystem, Listener Actor, and Listener Component.

Transcription Events¶

Core events for receiving transcription results.

Event	Parameters	Description
`OnUtteranceProcessed`	`FString TranscribedText`	The main event. Fires when a complete utterance has been transcribed. This is the final, confirmed text after VAD detects silence and the full audio segment is processed.
`OnSTTProcessingStarted`	`float UtteranceDuration`	Fires when audio is submitted to Whisper for transcription. `UtteranceDuration` is the length of the audio segment in seconds.
`OnSTTProcessingFailed`	`FString ErrorMessage`	Fires if transcription fails. Contains a human-readable error description.

Streaming Events¶

Events for real-time streaming transcription. Only fire when streaming is enabled.

Event	Parameters	Description
`OnTranscriptionUpdated`	`FString CommittedText`, `FString TentativeText`, `bool bTranscriptionFinished`	Ideal for live UI. Fires on every streaming snapshot. `CommittedText` contains words confirmed by Local Agreement. `TentativeText` contains unconfirmed words from the latest pass. `bTranscriptionFinished` is `true` on the final update.
`OnStreamingTextCommitted`	`FString CommittedText`, `FString AccumulatedText`	Fires when new words are confirmed by Local Agreement. `CommittedText` is the newly confirmed segment. `AccumulatedText` is all confirmed text so far.

Choosing a Streaming Event

Use OnTranscriptionUpdated for subtitle-style UI where you want to show both stable and in-progress text. Use OnStreamingTextCommitted when you only need confirmed words (e.g., feeding into a command parser).

VAD Events¶

Events driven by the Silero Voice Activity Detection model.

Event	Parameters	Description
`OnSpeechDetected`	`float InitialRMS`	Fires when the VAD transitions from silence to speech. `InitialRMS` is the audio level at the moment of detection (metering value, not a VAD input).
`OnSilenceDetected`	---	Fires when the VAD transitions from speech to silence after the configured minimum silence duration.
`OnVADUpdate`	`float CurrentRMS`, `bool bIsSpeech`	Periodic update with current audio level and VAD speech state. Useful for driving visual indicators.
`OnVADProbability`	`float Probability`, `bool bIsSpeech`	Raw Silero VAD output. Fires approximately 31 times per second (one per 512-sample window at 16kHz). `Probability` is the speech probability (0.0--1.0).
`OnUtteranceReady`	`int32 SampleCount`	Fires when a complete utterance has been captured and is ready for transcription. `SampleCount` is the total number of audio samples in the utterance.

System Events¶

Lifecycle and device-change notifications.

Event	Parameters	Description
`OnSTTSystemLoaded`	`bool bSuccess`	Fires after `LoadSTTSystem` completes. `bSuccess` indicates whether the model loaded successfully.
`OnPipelineStateChanged`	`EAbolethPipelineState OldState`, `EAbolethPipelineState NewState`	Fires on every pipeline state transition. See Structs & Enums for state values.
`OnAudioDeviceChanged`	`FAudioDeviceInfo NewDevice`	Fires when the active audio input device changes (via `SetAudioDeviceByIndex` or hardware hot-plug).
`OnAudioDevicesRefreshed`	---	Fires after `RefreshAudioDevices` completes enumeration.

Blueprint Binding¶

To bind events in Blueprint:

Place an AbolethListenerActor in the level or add an AbolethListenerComponent to an actor.
Select the actor/component and open the Details panel.
Scroll to the Events section.
Click the + button next to the event you want to handle (e.g., OnUtteranceProcessed).
The Event Graph opens with a new event node. Connect your logic to it.
For OnTranscriptionUpdated, wire CommittedText and TentativeText into a Text Block or subtitle widget for live display.

C++ Binding¶

Basic Transcription¶

void AMyActor::BeginPlay()
{
    Super::BeginPlay();

    UAbolethSTTSubsystem* STT = UAbolethSTTSubsystem::GetSTTSubsystem(this);
    if (!STT) return;

    STT->OnUtteranceProcessed.AddDynamic(this, &AMyActor::HandleUtterance);
    STT->OnSpeechDetected.AddDynamic(this, &AMyActor::HandleSpeechStart);
    STT->OnSilenceDetected.AddDynamic(this, &AMyActor::HandleSilenceStart);
    STT->OnPipelineStateChanged.AddDynamic(this, &AMyActor::HandleStateChange);

    STT->LoadSTTSystem();
    STT->StartListening();
}

void AMyActor::HandleUtterance(const FString& TranscribedText)
{
    UE_LOG(LogTemp, Log, TEXT("Player said: %s"), *TranscribedText);
    // Feed into dialogue system, command parser, etc.
}

void AMyActor::HandleSpeechStart(float InitialRMS)
{
    UE_LOG(LogTemp, Log, TEXT("Speech started (RMS: %.3f)"), InitialRMS);
    // Show "listening" indicator in UI
}

void AMyActor::HandleSilenceStart()
{
    UE_LOG(LogTemp, Log, TEXT("Silence detected"));
    // Hide "listening" indicator
}

void AMyActor::HandleStateChange(EAbolethPipelineState OldState,
                                  EAbolethPipelineState NewState)
{
    UE_LOG(LogTemp, Log, TEXT("Pipeline: %s -> %s"),
        *UEnum::GetValueAsString(OldState),
        *UEnum::GetValueAsString(NewState));
}

Live Streaming Text¶

void AMyActor::BeginPlay()
{
    Super::BeginPlay();

    UAbolethSTTSubsystem* STT = UAbolethSTTSubsystem::GetSTTSubsystem(this);
    if (!STT) return;

    STT->OnTranscriptionUpdated.AddDynamic(this, &AMyActor::HandleLiveText);

    STT->SetStreamingEnabled(true);
    STT->LoadSTTSystem();
    STT->StartListening();
}

void AMyActor::HandleLiveText(const FString& CommittedText,
                               const FString& TentativeText,
                               bool bTranscriptionFinished)
{
    // CommittedText  = confirmed words (stable, won't change)
    // TentativeText  = unconfirmed words (may change on next pass)

    FString DisplayText = CommittedText;
    if (!TentativeText.IsEmpty())
    {
        // Show tentative text in a different style (e.g., grey/italic)
        DisplayText += TEXT(" ") + TentativeText;
    }

    SubtitleWidget->SetText(FText::FromString(DisplayText));

    if (bTranscriptionFinished)
    {
        // Final result — tentative text is empty, committed is complete
        UE_LOG(LogTemp, Log, TEXT("Final: %s"), *CommittedText);
    }
}

Delegate Lifetime

Delegates bound via AddDynamic are automatically cleaned up when the bound object is destroyed. No manual unbinding is required in most cases.