Implement Phase 6R-H advanced narration orchestration
This commit is contained in:
axiomlogicnexus
parent
c25a2d7331
commit
3bede5500c
12 changed files with 738 additions and 12 deletions
|
|
@ -96,6 +96,42 @@ namespace HyperTwistContractLibraryInternal
|
|||
return Profile;
|
||||
}
|
||||
|
||||
FHyperTwistNarrationOrchestrationProfile MakeLocalNarrationOrchestrationProfile()
|
||||
{
|
||||
FHyperTwistNarrationOrchestrationProfile Profile;
|
||||
Profile.OrchestrationProfileId = TEXT("local-basic-narration-v1");
|
||||
Profile.ServiceLaneId = TEXT("voice/local-sidecar");
|
||||
Profile.ModelBindingId = TEXT("voice-model/local-offline-medium-v1");
|
||||
Profile.VocoderBindingId = TEXT("vocoder/default-mono-v1");
|
||||
Profile.DefaultSpeakerProfileId = TEXT("speaker/default");
|
||||
Profile.bEnableSpeakerSelection = false;
|
||||
Profile.bEnableLanguageRouting = false;
|
||||
Profile.bEnableSentenceSplitting = false;
|
||||
Profile.bEnableSpeakerReferenceConditioning = false;
|
||||
Profile.bSupportsMultiSpeaker = false;
|
||||
Profile.bSupportsMultiLingual = false;
|
||||
Profile.bPreferGpu = false;
|
||||
return Profile;
|
||||
}
|
||||
|
||||
FHyperTwistNarrationOrchestrationProfile MakeAdvancedNarrationOrchestrationProfile()
|
||||
{
|
||||
FHyperTwistNarrationOrchestrationProfile Profile;
|
||||
Profile.OrchestrationProfileId = TEXT("python-multispeaker-narration-v1");
|
||||
Profile.ServiceLaneId = TEXT("voice/python-service");
|
||||
Profile.ModelBindingId = TEXT("voice-model/multilingual-premium-v1");
|
||||
Profile.VocoderBindingId = TEXT("vocoder/high-fidelity-v1");
|
||||
Profile.DefaultSpeakerProfileId = TEXT("speaker/companion-guide");
|
||||
Profile.bEnableSpeakerSelection = true;
|
||||
Profile.bEnableLanguageRouting = true;
|
||||
Profile.bEnableSentenceSplitting = true;
|
||||
Profile.bEnableSpeakerReferenceConditioning = true;
|
||||
Profile.bSupportsMultiSpeaker = true;
|
||||
Profile.bSupportsMultiLingual = true;
|
||||
Profile.bPreferGpu = true;
|
||||
return Profile;
|
||||
}
|
||||
|
||||
FHyperTwistTrainingSourceAttribution MakeMethodDrillCleanRoomAttribution()
|
||||
{
|
||||
FHyperTwistTrainingSourceAttribution Attribution;
|
||||
|
|
@ -826,6 +862,18 @@ TArray<FHyperTwistVoiceProfileSummary> UHyperTwistContractLibrary::MakeSampleVoi
|
|||
1,
|
||||
TEXT("de/de_DE/thorsten/medium/MODEL_CARD"),
|
||||
{TEXT("de-de-thorsten")}
|
||||
),
|
||||
HyperTwistContractLibraryInternal::MakeVoiceProfile(
|
||||
TEXT("multilingual-guide-premium"),
|
||||
TEXT("Companion Guide"),
|
||||
TEXT("mul"),
|
||||
TEXT("mul"),
|
||||
TEXT("INTL"),
|
||||
TEXT("Multilingual"),
|
||||
TEXT("premium"),
|
||||
32,
|
||||
TEXT("mul/intl/companion-guide/premium/MODEL_CARD"),
|
||||
{TEXT("xtts-guide"), TEXT("companion-guide")}
|
||||
)
|
||||
};
|
||||
}
|
||||
|
|
@ -839,18 +887,46 @@ FHyperTwistNarrationSynthesisRequest UHyperTwistContractLibrary::MakeSampleNarra
|
|||
Request.OutputRouteId = TEXT("companion/narration/local");
|
||||
Request.VoiceProfileId = TEXT("en_US-lessac-medium");
|
||||
Request.LanguageCode = TEXT("en_US");
|
||||
Request.OrchestrationProfile = HyperTwistContractLibraryInternal::MakeLocalNarrationOrchestrationProfile();
|
||||
Request.ScriptText = TEXT("Good work. Keep your eyes on the next pair.");
|
||||
Request.SubtitleSeedText = Request.ScriptText;
|
||||
Request.SpeakerId = 0;
|
||||
Request.SpeakerProfileId = TEXT("speaker/default");
|
||||
Request.LengthScale = 1.0f;
|
||||
Request.NoiseScale = 0.667f;
|
||||
Request.NoiseW = 0.8f;
|
||||
Request.SentenceSilenceSeconds = 0.2f;
|
||||
Request.bSplitSentences = false;
|
||||
Request.bPreferCuda = false;
|
||||
Request.AudioEncodingHint = TEXT("wav-pcm16-mono");
|
||||
return Request;
|
||||
}
|
||||
|
||||
FHyperTwistNarrationSynthesisRequest UHyperTwistContractLibrary::MakeSampleAdvancedNarrationSynthesisRequest()
|
||||
{
|
||||
FHyperTwistNarrationSynthesisRequest Request;
|
||||
Request.RequestId = TEXT("narration_advanced_01");
|
||||
Request.NarrationContractId = TEXT("embodied-companion-narration");
|
||||
Request.ServiceLaneId = TEXT("voice/python-service");
|
||||
Request.OutputRouteId = TEXT("companion/narration/local");
|
||||
Request.VoiceProfileId = TEXT("multilingual-guide-premium");
|
||||
Request.LanguageCode = TEXT("fr_FR");
|
||||
Request.OrchestrationProfile = HyperTwistContractLibraryInternal::MakeAdvancedNarrationOrchestrationProfile();
|
||||
Request.ScriptText = TEXT("Bien joue. Garde les coins alignes pour le prochain exercice.");
|
||||
Request.SubtitleSeedText = Request.ScriptText;
|
||||
Request.SpeakerId = 7;
|
||||
Request.SpeakerProfileId = TEXT("speaker/companion-guide");
|
||||
Request.ReferenceVoiceSampleRef = TEXT("mock://voice/companion-guide/reference");
|
||||
Request.LengthScale = 0.95f;
|
||||
Request.NoiseScale = 0.5f;
|
||||
Request.NoiseW = 0.7f;
|
||||
Request.SentenceSilenceSeconds = 0.12f;
|
||||
Request.bSplitSentences = true;
|
||||
Request.bPreferCuda = true;
|
||||
Request.AudioEncodingHint = TEXT("wav-pcm16-mono");
|
||||
return Request;
|
||||
}
|
||||
|
||||
FHyperTwistNarrationSynthesisResult UHyperTwistContractLibrary::MakeMockNarrationSynthesisResult()
|
||||
{
|
||||
const FHyperTwistNarrationSynthesisRequest Request = MakeSampleNarrationSynthesisRequest();
|
||||
|
|
@ -862,6 +938,12 @@ FHyperTwistNarrationSynthesisResult UHyperTwistContractLibrary::MakeMockNarratio
|
|||
Result.OutputRouteId = Request.OutputRouteId;
|
||||
Result.VoiceProfileId = Request.VoiceProfileId;
|
||||
Result.LanguageCode = Request.LanguageCode;
|
||||
Result.OrchestrationProfileId = Request.OrchestrationProfile.OrchestrationProfileId;
|
||||
Result.ModelBindingId = Request.OrchestrationProfile.ModelBindingId;
|
||||
Result.VocoderBindingId = Request.OrchestrationProfile.VocoderBindingId;
|
||||
Result.SpeakerProfileId = Request.SpeakerProfileId.IsEmpty()
|
||||
? Request.OrchestrationProfile.DefaultSpeakerProfileId
|
||||
: Request.SpeakerProfileId;
|
||||
Result.SubtitleText = Request.SubtitleSeedText;
|
||||
Result.AudioEncoding = TEXT("wav-pcm16-mono");
|
||||
Result.SampleRateHz = 16000;
|
||||
|
|
@ -871,6 +953,11 @@ FHyperTwistNarrationSynthesisResult UHyperTwistContractLibrary::MakeMockNarratio
|
|||
Result.AppliedNoiseScale = Request.NoiseScale;
|
||||
Result.AppliedNoiseW = Request.NoiseW;
|
||||
Result.AppliedSentenceSilenceSeconds = Request.SentenceSilenceSeconds;
|
||||
Result.bUsedSentenceSplitting = Request.bSplitSentences || Request.OrchestrationProfile.bEnableSentenceSplitting;
|
||||
Result.bUsedSpeakerSelection =
|
||||
Request.OrchestrationProfile.bEnableSpeakerSelection && !Result.SpeakerProfileId.IsEmpty();
|
||||
Result.bUsedSpeakerReferenceConditioning =
|
||||
Request.OrchestrationProfile.bEnableSpeakerReferenceConditioning && !Request.ReferenceVoiceSampleRef.IsEmpty();
|
||||
Result.AudioBytes = HyperTwistContractLibraryInternal::MakeSilentMonoPcm16Wav(
|
||||
Result.SampleRateHz,
|
||||
Result.DurationMs
|
||||
|
|
@ -889,7 +976,13 @@ FHyperTwistVoiceServiceHealth UHyperTwistContractLibrary::MakeMockVoiceServiceHe
|
|||
TEXT("offlineNarration"),
|
||||
TEXT("wavOutput"),
|
||||
TEXT("subtitleSeed"),
|
||||
TEXT("localProfileCatalog")
|
||||
TEXT("localProfileCatalog"),
|
||||
TEXT("multiSpeakerNarration"),
|
||||
TEXT("multiLingualNarration"),
|
||||
TEXT("speakerReferenceConditioning"),
|
||||
TEXT("sentenceSplitNarration"),
|
||||
TEXT("modelBindingSelection"),
|
||||
TEXT("vocoderBindingSelection")
|
||||
};
|
||||
|
||||
for (const FHyperTwistVoiceProfileSummary& Profile : MakeSampleVoiceProfiles())
|
||||
|
|
@ -897,6 +990,19 @@ FHyperTwistVoiceServiceHealth UHyperTwistContractLibrary::MakeMockVoiceServiceHe
|
|||
Health.SupportedVoiceProfileIds.Add(Profile.VoiceProfileId);
|
||||
}
|
||||
|
||||
Health.SupportedOrchestrationProfiles = {
|
||||
TEXT("local-basic-narration-v1"),
|
||||
TEXT("python-multispeaker-narration-v1")
|
||||
};
|
||||
Health.SupportedModelBindingIds = {
|
||||
TEXT("voice-model/local-offline-medium-v1"),
|
||||
TEXT("voice-model/multilingual-premium-v1")
|
||||
};
|
||||
Health.SupportedVocoderBindingIds = {
|
||||
TEXT("vocoder/default-mono-v1"),
|
||||
TEXT("vocoder/high-fidelity-v1")
|
||||
};
|
||||
|
||||
Health.bReady = true;
|
||||
return Health;
|
||||
}
|
||||
|
|
|
|||
|
|
@ -309,12 +309,23 @@ FHyperTwistNarrationSynthesisResult UHyperTwistHttpVoiceClient::SynthesizeNarrat
|
|||
Result.OutputRouteId = Request.OutputRouteId;
|
||||
Result.VoiceProfileId = Request.VoiceProfileId;
|
||||
Result.LanguageCode = Request.LanguageCode;
|
||||
Result.OrchestrationProfileId = Request.OrchestrationProfile.OrchestrationProfileId;
|
||||
Result.ModelBindingId = Request.OrchestrationProfile.ModelBindingId;
|
||||
Result.VocoderBindingId = Request.OrchestrationProfile.VocoderBindingId;
|
||||
Result.SpeakerProfileId = Request.SpeakerProfileId.IsEmpty()
|
||||
? Request.OrchestrationProfile.DefaultSpeakerProfileId
|
||||
: Request.SpeakerProfileId;
|
||||
Result.SubtitleText = Request.SubtitleSeedText.IsEmpty() ? Request.ScriptText : Request.SubtitleSeedText;
|
||||
Result.AudioEncoding = TEXT("wav-pcm16-mono");
|
||||
Result.AppliedLengthScale = Request.LengthScale;
|
||||
Result.AppliedNoiseScale = Request.NoiseScale;
|
||||
Result.AppliedNoiseW = Request.NoiseW;
|
||||
Result.AppliedSentenceSilenceSeconds = Request.SentenceSilenceSeconds;
|
||||
Result.bUsedSentenceSplitting = Request.bSplitSentences || Request.OrchestrationProfile.bEnableSentenceSplitting;
|
||||
Result.bUsedSpeakerSelection =
|
||||
Request.OrchestrationProfile.bEnableSpeakerSelection && !Result.SpeakerProfileId.IsEmpty();
|
||||
Result.bUsedSpeakerReferenceConditioning =
|
||||
Request.OrchestrationProfile.bEnableSpeakerReferenceConditioning && !Request.ReferenceVoiceSampleRef.IsEmpty();
|
||||
|
||||
if (!Request.IsStructurallyValid())
|
||||
{
|
||||
|
|
@ -379,7 +390,13 @@ FHyperTwistVoiceServiceHealth UHyperTwistHttpVoiceClient::GetVoiceServiceHealth(
|
|||
TEXT("voiceCatalog"),
|
||||
TEXT("offlineNarration"),
|
||||
TEXT("wavOutput"),
|
||||
TEXT("subtitleSeed")
|
||||
TEXT("subtitleSeed"),
|
||||
TEXT("multiSpeakerNarration"),
|
||||
TEXT("multiLingualNarration"),
|
||||
TEXT("speakerReferenceConditioning"),
|
||||
TEXT("sentenceSplitNarration"),
|
||||
TEXT("modelBindingSelection"),
|
||||
TEXT("vocoderBindingSelection")
|
||||
};
|
||||
Health.bReady = false;
|
||||
|
||||
|
|
@ -412,7 +429,25 @@ FHyperTwistVoiceServiceHealth UHyperTwistHttpVoiceClient::GetVoiceServiceHealth(
|
|||
TEXT("voiceCatalog"),
|
||||
TEXT("offlineNarration"),
|
||||
TEXT("wavOutput"),
|
||||
TEXT("subtitleSeed")
|
||||
TEXT("subtitleSeed"),
|
||||
TEXT("multiSpeakerNarration"),
|
||||
TEXT("multiLingualNarration"),
|
||||
TEXT("speakerReferenceConditioning"),
|
||||
TEXT("sentenceSplitNarration"),
|
||||
TEXT("modelBindingSelection"),
|
||||
TEXT("vocoderBindingSelection")
|
||||
};
|
||||
Health.SupportedOrchestrationProfiles = {
|
||||
TEXT("local-basic-narration-v1"),
|
||||
TEXT("python-multispeaker-narration-v1")
|
||||
};
|
||||
Health.SupportedModelBindingIds = {
|
||||
TEXT("voice-model/local-offline-medium-v1"),
|
||||
TEXT("voice-model/multilingual-premium-v1")
|
||||
};
|
||||
Health.SupportedVocoderBindingIds = {
|
||||
TEXT("vocoder/default-mono-v1"),
|
||||
TEXT("vocoder/high-fidelity-v1")
|
||||
};
|
||||
Health.bReady = false;
|
||||
Health.LastError = LastTransportError;
|
||||
|
|
@ -434,6 +469,18 @@ FHyperTwistVoiceServiceHealth UHyperTwistHttpVoiceClient::GetVoiceServiceHealth(
|
|||
Health.Capabilities.AddUnique(TEXT("offlineNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("wavOutput"));
|
||||
Health.Capabilities.AddUnique(TEXT("subtitleSeed"));
|
||||
Health.Capabilities.AddUnique(TEXT("multiSpeakerNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("multiLingualNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("speakerReferenceConditioning"));
|
||||
Health.Capabilities.AddUnique(TEXT("sentenceSplitNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("modelBindingSelection"));
|
||||
Health.Capabilities.AddUnique(TEXT("vocoderBindingSelection"));
|
||||
Health.SupportedOrchestrationProfiles.AddUnique(TEXT("local-basic-narration-v1"));
|
||||
Health.SupportedOrchestrationProfiles.AddUnique(TEXT("python-multispeaker-narration-v1"));
|
||||
Health.SupportedModelBindingIds.AddUnique(TEXT("voice-model/local-offline-medium-v1"));
|
||||
Health.SupportedModelBindingIds.AddUnique(TEXT("voice-model/multilingual-premium-v1"));
|
||||
Health.SupportedVocoderBindingIds.AddUnique(TEXT("vocoder/default-mono-v1"));
|
||||
Health.SupportedVocoderBindingIds.AddUnique(TEXT("vocoder/high-fidelity-v1"));
|
||||
LastTransportError.Reset();
|
||||
Health.LastError.Reset();
|
||||
return Health;
|
||||
|
|
|
|||
|
|
@ -18,11 +18,22 @@ FHyperTwistNarrationSynthesisResult UHyperTwistMockVoiceClient::SynthesizeNarrat
|
|||
Result.OutputRouteId = Request.OutputRouteId;
|
||||
Result.VoiceProfileId = Request.VoiceProfileId;
|
||||
Result.LanguageCode = Request.LanguageCode;
|
||||
Result.OrchestrationProfileId = Request.OrchestrationProfile.OrchestrationProfileId;
|
||||
Result.ModelBindingId = Request.OrchestrationProfile.ModelBindingId;
|
||||
Result.VocoderBindingId = Request.OrchestrationProfile.VocoderBindingId;
|
||||
Result.SpeakerProfileId = Request.SpeakerProfileId.IsEmpty()
|
||||
? Request.OrchestrationProfile.DefaultSpeakerProfileId
|
||||
: Request.SpeakerProfileId;
|
||||
Result.SubtitleText = Request.SubtitleSeedText.IsEmpty() ? Request.ScriptText : Request.SubtitleSeedText;
|
||||
Result.AppliedLengthScale = Request.LengthScale;
|
||||
Result.AppliedNoiseScale = Request.NoiseScale;
|
||||
Result.AppliedNoiseW = Request.NoiseW;
|
||||
Result.AppliedSentenceSilenceSeconds = Request.SentenceSilenceSeconds;
|
||||
Result.bUsedSentenceSplitting = Request.bSplitSentences || Request.OrchestrationProfile.bEnableSentenceSplitting;
|
||||
Result.bUsedSpeakerSelection =
|
||||
Request.OrchestrationProfile.bEnableSpeakerSelection && !Result.SpeakerProfileId.IsEmpty();
|
||||
Result.bUsedSpeakerReferenceConditioning =
|
||||
Request.OrchestrationProfile.bEnableSpeakerReferenceConditioning && !Request.ReferenceVoiceSampleRef.IsEmpty();
|
||||
|
||||
if (!Request.IsStructurallyValid())
|
||||
{
|
||||
|
|
@ -33,6 +44,12 @@ FHyperTwistNarrationSynthesisResult UHyperTwistMockVoiceClient::SynthesizeNarrat
|
|||
Result.OutputRouteId = Request.OutputRouteId;
|
||||
Result.VoiceProfileId = Request.VoiceProfileId;
|
||||
Result.LanguageCode = Request.LanguageCode;
|
||||
Result.OrchestrationProfileId = Request.OrchestrationProfile.OrchestrationProfileId;
|
||||
Result.ModelBindingId = Request.OrchestrationProfile.ModelBindingId;
|
||||
Result.VocoderBindingId = Request.OrchestrationProfile.VocoderBindingId;
|
||||
Result.SpeakerProfileId = Request.SpeakerProfileId.IsEmpty()
|
||||
? Request.OrchestrationProfile.DefaultSpeakerProfileId
|
||||
: Request.SpeakerProfileId;
|
||||
Result.Warnings.Add(TEXT("narration-request-invalid"));
|
||||
return Result;
|
||||
}
|
||||
|
|
@ -64,5 +81,13 @@ FHyperTwistVoiceServiceHealth UHyperTwistMockVoiceClient::GetVoiceServiceHealth(
|
|||
Health.Capabilities.AddUnique(TEXT("voiceCatalog"));
|
||||
Health.Capabilities.AddUnique(TEXT("wavOutput"));
|
||||
Health.Capabilities.AddUnique(TEXT("subtitleSeed"));
|
||||
Health.Capabilities.AddUnique(TEXT("multiSpeakerNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("multiLingualNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("speakerReferenceConditioning"));
|
||||
Health.Capabilities.AddUnique(TEXT("sentenceSplitNarration"));
|
||||
Health.Capabilities.AddUnique(TEXT("modelBindingSelection"));
|
||||
Health.Capabilities.AddUnique(TEXT("vocoderBindingSelection"));
|
||||
Health.SupportedOrchestrationProfiles.AddUnique(TEXT("local-basic-narration-v1"));
|
||||
Health.SupportedOrchestrationProfiles.AddUnique(TEXT("python-multispeaker-narration-v1"));
|
||||
return Health;
|
||||
}
|
||||
|
|
|
|||
|
|
@ -72,6 +72,9 @@ public:
|
|||
UFUNCTION(BlueprintPure, Category = "HyperTwist|Recognition")
|
||||
static FHyperTwistNarrationSynthesisRequest MakeSampleNarrationSynthesisRequest();
|
||||
|
||||
UFUNCTION(BlueprintPure, Category = "HyperTwist|Recognition")
|
||||
static FHyperTwistNarrationSynthesisRequest MakeSampleAdvancedNarrationSynthesisRequest();
|
||||
|
||||
UFUNCTION(BlueprintPure, Category = "HyperTwist|Recognition")
|
||||
static FHyperTwistNarrationSynthesisResult MakeMockNarrationSynthesisResult();
|
||||
|
||||
|
|
|
|||
|
|
@ -1006,6 +1006,57 @@ struct FHyperTwistVoiceProfileCatalog
|
|||
}
|
||||
};
|
||||
|
||||
USTRUCT(BlueprintType)
|
||||
struct FHyperTwistNarrationOrchestrationProfile
|
||||
{
|
||||
GENERATED_BODY()
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString OrchestrationProfileId = TEXT("local-basic-narration-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString ServiceLaneId = TEXT("voice/local-sidecar");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString ModelBindingId = TEXT("voice-model/local-offline-medium-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString VocoderBindingId = TEXT("vocoder/default-mono-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString DefaultSpeakerProfileId = TEXT("speaker/default");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bEnableSpeakerSelection = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bEnableLanguageRouting = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bEnableSentenceSplitting = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bEnableSpeakerReferenceConditioning = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bSupportsMultiSpeaker = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bSupportsMultiLingual = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bPreferGpu = false;
|
||||
|
||||
bool IsStructurallyValid() const
|
||||
{
|
||||
return !OrchestrationProfileId.IsEmpty()
|
||||
&& !ServiceLaneId.IsEmpty()
|
||||
&& !ModelBindingId.IsEmpty()
|
||||
&& !VocoderBindingId.IsEmpty()
|
||||
&& !DefaultSpeakerProfileId.IsEmpty();
|
||||
}
|
||||
};
|
||||
|
||||
USTRUCT(BlueprintType)
|
||||
struct FHyperTwistNarrationSynthesisRequest
|
||||
{
|
||||
|
|
@ -1029,6 +1080,9 @@ struct FHyperTwistNarrationSynthesisRequest
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString LanguageCode = TEXT("en_US");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FHyperTwistNarrationOrchestrationProfile OrchestrationProfile;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString ScriptText;
|
||||
|
||||
|
|
@ -1038,6 +1092,12 @@ struct FHyperTwistNarrationSynthesisRequest
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
int32 SpeakerId = 0;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString SpeakerProfileId;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString ReferenceVoiceSampleRef;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
float LengthScale = 1.0f;
|
||||
|
||||
|
|
@ -1050,6 +1110,9 @@ struct FHyperTwistNarrationSynthesisRequest
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
float SentenceSilenceSeconds = 0.2f;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bSplitSentences = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bPreferCuda = false;
|
||||
|
||||
|
|
@ -1064,6 +1127,8 @@ struct FHyperTwistNarrationSynthesisRequest
|
|||
&& !OutputRouteId.IsEmpty()
|
||||
&& !VoiceProfileId.IsEmpty()
|
||||
&& !LanguageCode.IsEmpty()
|
||||
&& OrchestrationProfile.IsStructurallyValid()
|
||||
&& ServiceLaneId.Equals(OrchestrationProfile.ServiceLaneId, ESearchCase::CaseSensitive)
|
||||
&& !ScriptText.IsEmpty()
|
||||
&& !AudioEncodingHint.IsEmpty()
|
||||
&& LengthScale > 0.0f
|
||||
|
|
@ -1096,6 +1161,18 @@ struct FHyperTwistNarrationSynthesisResult
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString LanguageCode = TEXT("en_US");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString OrchestrationProfileId = TEXT("local-basic-narration-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString ModelBindingId = TEXT("voice-model/local-offline-medium-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString VocoderBindingId = TEXT("vocoder/default-mono-v1");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString SpeakerProfileId = TEXT("speaker/default");
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString SubtitleText;
|
||||
|
||||
|
|
@ -1123,6 +1200,15 @@ struct FHyperTwistNarrationSynthesisResult
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
float AppliedSentenceSilenceSeconds = 0.2f;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bUsedSentenceSplitting = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bUsedSpeakerSelection = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
bool bUsedSpeakerReferenceConditioning = false;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
TArray<uint8> AudioBytes;
|
||||
|
||||
|
|
@ -1137,6 +1223,9 @@ struct FHyperTwistNarrationSynthesisResult
|
|||
&& !OutputRouteId.IsEmpty()
|
||||
&& !VoiceProfileId.IsEmpty()
|
||||
&& !LanguageCode.IsEmpty()
|
||||
&& !OrchestrationProfileId.IsEmpty()
|
||||
&& !ModelBindingId.IsEmpty()
|
||||
&& !VocoderBindingId.IsEmpty()
|
||||
&& !SubtitleText.IsEmpty()
|
||||
&& !AudioEncoding.IsEmpty()
|
||||
&& SampleRateHz > 0
|
||||
|
|
@ -1170,6 +1259,15 @@ struct FHyperTwistVoiceServiceHealth
|
|||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
TArray<FString> SupportedVoiceProfileIds;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
TArray<FString> SupportedOrchestrationProfiles;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
TArray<FString> SupportedModelBindingIds;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
TArray<FString> SupportedVocoderBindingIds;
|
||||
|
||||
UPROPERTY(EditAnywhere, BlueprintReadWrite, Category = "HyperTwist")
|
||||
FString LastError;
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,189 @@
|
|||
// Copyright HyperTwist, Inc. All Rights Reserved.
|
||||
|
||||
#include "Misc/AutomationTest.h"
|
||||
|
||||
#include "HyperTwistBootstrap/HyperTwistContractLibrary.h"
|
||||
#include "HyperTwistRecognition/HyperTwistVoiceClient.h"
|
||||
#include "HyperTwistTraining/HyperTwistTrainingRuntimeLibrary.h"
|
||||
|
||||
#if WITH_AUTOMATION_TESTS
|
||||
|
||||
namespace HyperTwistCoquiPhase6RHTestInternal
|
||||
{
|
||||
UHyperTwistMockVoiceClient* MakeMockVoiceClient()
|
||||
{
|
||||
return NewObject<UHyperTwistMockVoiceClient>(GetTransientPackage());
|
||||
}
|
||||
}
|
||||
|
||||
IMPLEMENT_SIMPLE_AUTOMATION_TEST(
|
||||
FHyperTwistCoquiPhase6RHAdvancedNarrationRequestTest,
|
||||
"HyperTwist.Permissive.Coqui.Phase6R.H.AdvancedNarrationRequest",
|
||||
EAutomationTestFlags::EditorContext | EAutomationTestFlags::EngineFilter
|
||||
)
|
||||
|
||||
bool FHyperTwistCoquiPhase6RHAdvancedNarrationRequestTest::RunTest(const FString& Parameters)
|
||||
{
|
||||
FHyperTwistTrainingCompanionNarrationContract NarrationContract;
|
||||
TestTrue(
|
||||
TEXT("The bundled embodied companion narration contract must stay available for Phase 6R-H."),
|
||||
UHyperTwistTrainingRuntimeLibrary::TryGetBundledEmbodiedCompanionNarrationContract(
|
||||
TEXT("embodied-companion-narration"),
|
||||
NarrationContract
|
||||
)
|
||||
);
|
||||
TestTrue(TEXT("The bundled narration contract must remain structurally valid."), NarrationContract.IsStructurallyValid());
|
||||
|
||||
UHyperTwistMockVoiceClient* VoiceClient = HyperTwistCoquiPhase6RHTestInternal::MakeMockVoiceClient();
|
||||
TestNotNull(TEXT("The mock voice client must be constructed for Phase 6R-H."), VoiceClient);
|
||||
if (VoiceClient == nullptr)
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
const TArray<FHyperTwistVoiceProfileSummary> Profiles = VoiceClient->ListVoiceProfiles();
|
||||
TestTrue(TEXT("The retained advanced narration slice must expose a multilingual guide profile."), Profiles.ContainsByPredicate(
|
||||
[](const FHyperTwistVoiceProfileSummary& Profile)
|
||||
{
|
||||
return Profile.VoiceProfileId == TEXT("multilingual-guide-premium")
|
||||
&& Profile.NumSpeakers >= 2;
|
||||
}
|
||||
));
|
||||
|
||||
const FHyperTwistNarrationSynthesisRequest Request =
|
||||
UHyperTwistContractLibrary::MakeSampleAdvancedNarrationSynthesisRequest();
|
||||
TestTrue(TEXT("The advanced narration request must be structurally valid."), Request.IsStructurallyValid());
|
||||
TestEqual(
|
||||
TEXT("The advanced request must stay on the embodied companion narration contract."),
|
||||
Request.NarrationContractId,
|
||||
TEXT("embodied-companion-narration")
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced request must move to the Python voice service lane."),
|
||||
Request.ServiceLaneId,
|
||||
TEXT("voice/python-service")
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced orchestration profile must be bounded to the retained multispeaker narration slice."),
|
||||
Request.OrchestrationProfile.OrchestrationProfileId,
|
||||
TEXT("python-multispeaker-narration-v1")
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced request must carry the normalized narration model binding."),
|
||||
Request.OrchestrationProfile.ModelBindingId,
|
||||
TEXT("voice-model/multilingual-premium-v1")
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced request must carry the normalized vocoder binding."),
|
||||
Request.OrchestrationProfile.VocoderBindingId,
|
||||
TEXT("vocoder/high-fidelity-v1")
|
||||
);
|
||||
TestTrue(TEXT("The advanced request must enable speaker selection."), Request.OrchestrationProfile.bEnableSpeakerSelection);
|
||||
TestTrue(TEXT("The advanced request must enable language routing."), Request.OrchestrationProfile.bEnableLanguageRouting);
|
||||
TestTrue(TEXT("The advanced request must enable sentence splitting."), Request.bSplitSentences);
|
||||
TestTrue(
|
||||
TEXT("The advanced request must retain a reference-voice sample ref without making voice assets part of the packet."),
|
||||
!Request.ReferenceVoiceSampleRef.IsEmpty()
|
||||
);
|
||||
return true;
|
||||
}
|
||||
|
||||
IMPLEMENT_SIMPLE_AUTOMATION_TEST(
|
||||
FHyperTwistCoquiPhase6RHAdvancedNarrationResultTest,
|
||||
"HyperTwist.Permissive.Coqui.Phase6R.H.AdvancedNarrationResult",
|
||||
EAutomationTestFlags::EditorContext | EAutomationTestFlags::EngineFilter
|
||||
)
|
||||
|
||||
bool FHyperTwistCoquiPhase6RHAdvancedNarrationResultTest::RunTest(const FString& Parameters)
|
||||
{
|
||||
UHyperTwistMockVoiceClient* VoiceClient = HyperTwistCoquiPhase6RHTestInternal::MakeMockVoiceClient();
|
||||
TestNotNull(TEXT("The mock voice client must be constructed for Phase 6R-H."), VoiceClient);
|
||||
if (VoiceClient == nullptr)
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
const FHyperTwistNarrationSynthesisRequest Request =
|
||||
UHyperTwistContractLibrary::MakeSampleAdvancedNarrationSynthesisRequest();
|
||||
const FHyperTwistNarrationSynthesisResult Result = VoiceClient->SynthesizeNarration(Request);
|
||||
|
||||
TestTrue(TEXT("The advanced narration synthesis result must stay structurally valid."), Result.IsStructurallyValid());
|
||||
TestEqual(TEXT("The advanced result must preserve the request id."), Result.RequestId, Request.RequestId);
|
||||
TestEqual(
|
||||
TEXT("The advanced result must preserve the advanced orchestration profile id."),
|
||||
Result.OrchestrationProfileId,
|
||||
Request.OrchestrationProfile.OrchestrationProfileId
|
||||
);
|
||||
TestEqual(TEXT("The advanced result must stay on the Python voice service lane."), Result.ServiceLaneId, TEXT("voice/python-service"));
|
||||
TestEqual(
|
||||
TEXT("The advanced result must preserve the normalized narration model binding."),
|
||||
Result.ModelBindingId,
|
||||
Request.OrchestrationProfile.ModelBindingId
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced result must preserve the normalized vocoder binding."),
|
||||
Result.VocoderBindingId,
|
||||
Request.OrchestrationProfile.VocoderBindingId
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced result must preserve the selected speaker profile."),
|
||||
Result.SpeakerProfileId,
|
||||
Request.SpeakerProfileId
|
||||
);
|
||||
TestTrue(TEXT("The advanced result must report sentence splitting usage."), Result.bUsedSentenceSplitting);
|
||||
TestTrue(TEXT("The advanced result must report speaker selection usage."), Result.bUsedSpeakerSelection);
|
||||
TestTrue(
|
||||
TEXT("The advanced result must report reference-voice conditioning usage."),
|
||||
Result.bUsedSpeakerReferenceConditioning
|
||||
);
|
||||
TestEqual(
|
||||
TEXT("The advanced result must preserve subtitle text."),
|
||||
Result.SubtitleText,
|
||||
Request.SubtitleSeedText
|
||||
);
|
||||
return true;
|
||||
}
|
||||
|
||||
IMPLEMENT_SIMPLE_AUTOMATION_TEST(
|
||||
FHyperTwistCoquiPhase6RHServiceHealthTest,
|
||||
"HyperTwist.Permissive.Coqui.Phase6R.H.ServiceHealth",
|
||||
EAutomationTestFlags::EditorContext | EAutomationTestFlags::EngineFilter
|
||||
)
|
||||
|
||||
bool FHyperTwistCoquiPhase6RHServiceHealthTest::RunTest(const FString& Parameters)
|
||||
{
|
||||
UHyperTwistMockVoiceClient* VoiceClient = HyperTwistCoquiPhase6RHTestInternal::MakeMockVoiceClient();
|
||||
TestNotNull(TEXT("The mock voice client must be constructed for Phase 6R-H."), VoiceClient);
|
||||
if (VoiceClient == nullptr)
|
||||
{
|
||||
return false;
|
||||
}
|
||||
|
||||
const FHyperTwistVoiceServiceHealth Health = VoiceClient->GetVoiceServiceHealth();
|
||||
TestTrue(TEXT("The advanced voice service health must be structurally valid."), Health.IsStructurallyValid());
|
||||
TestTrue(TEXT("The advanced voice service health must report ready."), Health.bReady);
|
||||
TestTrue(TEXT("The advanced voice service must expose multispeaker narration capability."), Health.Capabilities.Contains(TEXT("multiSpeakerNarration")));
|
||||
TestTrue(TEXT("The advanced voice service must expose multilingual narration capability."), Health.Capabilities.Contains(TEXT("multiLingualNarration")));
|
||||
TestTrue(
|
||||
TEXT("The advanced voice service must expose speaker-reference conditioning capability."),
|
||||
Health.Capabilities.Contains(TEXT("speakerReferenceConditioning"))
|
||||
);
|
||||
TestTrue(TEXT("The advanced voice service must expose sentence-split narration capability."), Health.Capabilities.Contains(TEXT("sentenceSplitNarration")));
|
||||
TestTrue(TEXT("The advanced voice service must expose model-binding selection capability."), Health.Capabilities.Contains(TEXT("modelBindingSelection")));
|
||||
TestTrue(TEXT("The advanced voice service must expose vocoder-binding selection capability."), Health.Capabilities.Contains(TEXT("vocoderBindingSelection")));
|
||||
TestTrue(
|
||||
TEXT("The advanced voice service must advertise the retained advanced orchestration profile."),
|
||||
Health.SupportedOrchestrationProfiles.Contains(TEXT("python-multispeaker-narration-v1"))
|
||||
);
|
||||
TestTrue(
|
||||
TEXT("The advanced voice service must advertise the normalized premium narration model binding."),
|
||||
Health.SupportedModelBindingIds.Contains(TEXT("voice-model/multilingual-premium-v1"))
|
||||
);
|
||||
TestTrue(
|
||||
TEXT("The advanced voice service must advertise the normalized high-fidelity vocoder binding."),
|
||||
Health.SupportedVocoderBindingIds.Contains(TEXT("vocoder/high-fidelity-v1"))
|
||||
);
|
||||
return true;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
|
@ -1312,6 +1312,7 @@ Decision date:
|
|||
|
||||
- `2026-04-24`
|
||||
- refreshed on `2026-05-13`
|
||||
- refreshed on `2026-05-22`
|
||||
|
||||
Current licensing judgment:
|
||||
|
||||
|
|
@ -1343,6 +1344,15 @@ Approved working posture:
|
|||
- do not collapse code and model-weight licensing into one bucket
|
||||
- track each chosen voice model or downloadable payload separately at implementation time
|
||||
- prefer a bounded speech sidecar or service seam so the product can swap or isolate model choices later without entangling the core coaching logic
|
||||
- the first bounded advanced narration slice is now landed:
|
||||
- advanced narration orchestration profile
|
||||
- normalized model/vocoder binding metadata
|
||||
- multilingual routing and speaker-selection flags
|
||||
- reference-speaker conditioning flag
|
||||
- bounded mock and HTTP voice-client seam support for the richer narration profile
|
||||
- advanced voice service-health capability exposure
|
||||
- `coqui-ai/TTS` remains a bounded donor, not the broad voice-platform owner
|
||||
- keep downloadable voice/model review separate from the code-license judgment
|
||||
|
||||
### `rhasspy/piper`
|
||||
|
||||
|
|
|
|||
|
|
@ -0,0 +1,144 @@
|
|||
# HyperTwist Phase 6R-H coqui-ai/TTS advanced narration implementation packet
|
||||
|
||||
Created on `2026-05-22`
|
||||
|
||||
## Status
|
||||
|
||||
- first-party HyperTwist packet
|
||||
- bounded permissive `Phase 6R-H` implementation slice
|
||||
|
||||
## Purpose
|
||||
|
||||
This packet lands the first narrower bounded slice from the retained `coqui-ai/TTS` voice-output
|
||||
row.
|
||||
|
||||
The landed slice is:
|
||||
|
||||
- first-party advanced narration orchestration profile above the existing embodied companion
|
||||
narration lane
|
||||
|
||||
It is not:
|
||||
|
||||
- a broad TTS platform packet
|
||||
- a downloadable voice-asset or model-shipping packet
|
||||
- a provider-profile or BYOK packet
|
||||
- a voice-conversion packet
|
||||
- a viseme / gesture runtime packet
|
||||
|
||||
## Current authority basis
|
||||
|
||||
This implementation packet stands on:
|
||||
|
||||
- `docs/HYPERTWIST_PHASE_0R_PACKET_0R_A_EVALUATION_2026-05-12.md`
|
||||
- `docs/HYPERTWIST_PHASE_2R_PACKET_2R_A_OWNERSHIP_AND_ACCEPTANCE_CONTRACT_2026-05-13.md`
|
||||
- `docs/REPO_LICENSE_TRACKING.md`
|
||||
- `docs/ops/HYPERTWIST_PROVIDER_NEUTRALITY_AND_BYOK_DOCTRINE_2026-05-21.md`
|
||||
- `docs/arch/HYPERTWIST_PHASE6R_H_COQUI_TTS_ADVANCED_NARRATION_PREPARATION_PACKET_2026-05-22.md`
|
||||
|
||||
The retained owner remains:
|
||||
|
||||
- `coqui-ai/TTS`
|
||||
|
||||
The top-level provider and companion contract owner does not change here:
|
||||
|
||||
- first-party HyperTwist remains the provider-neutral and companion-contract owner
|
||||
- this packet keeps `coqui-ai/TTS` bounded to the narrower `A1 / R2 / F2` advanced narration
|
||||
orchestration slice
|
||||
|
||||
## Landed scope
|
||||
|
||||
The current code now owns a retained advanced narration orchestration contract through:
|
||||
|
||||
- retained recognition contract types for:
|
||||
- narration orchestration profile
|
||||
- advanced narration request metadata
|
||||
- advanced narration result metadata
|
||||
- advanced voice service health metadata
|
||||
- sample voice profiles, basic and advanced narration request shaping, narration result shaping, and
|
||||
voice service health outputs in:
|
||||
- `UHyperTwistContractLibrary`
|
||||
- bounded voice client seam support in:
|
||||
- `UHyperTwistMockVoiceClient`
|
||||
- `UHyperTwistHttpVoiceClient`
|
||||
- focused automation coverage in:
|
||||
- `HyperTwistCoquiPhase6RHAdvancedNarrationContractTest.cpp`
|
||||
|
||||
## Why this is still intentionally bounded
|
||||
|
||||
This packet lands a richer voice-output donor seam, but it does not widen into the neighboring
|
||||
retained families.
|
||||
|
||||
Still deferred:
|
||||
|
||||
- downloadable voice/model review
|
||||
- bundled voice asset shipping
|
||||
- voice conversion
|
||||
- broad playback runtime ownership
|
||||
- viseme / gesture runtime integration
|
||||
- provider-profile and BYOK custody
|
||||
- broad assistant-platform scope
|
||||
|
||||
## Validation
|
||||
|
||||
Build validation:
|
||||
|
||||
- `Build.bat UnrealHyperTwistEditor Win64 Development -Project='C:\HyperTwist\UnrealHyperTwist\UnrealHyperTwist.uproject' -WaitMutex -NoHotReloadFromIDE`
|
||||
|
||||
Focused automation validation:
|
||||
|
||||
- `Automation RunTests HyperTwist.Permissive.Coqui.Phase6R.H`
|
||||
|
||||
Regression automation validation:
|
||||
|
||||
- `Automation RunTests HyperTwist.Permissive.Piper.Phase6R.G`
|
||||
- `Automation RunTests HyperTwist.Permissive.FasterWhisper.Phase6R.F`
|
||||
- `Automation RunTests HyperTwist.Permissive.WhisperCpp.Phase6R.E`
|
||||
- `Automation RunTests HyperTwist.CleanRoom.CubeDesk`
|
||||
|
||||
Expected covered tests:
|
||||
|
||||
- `AdvancedNarrationRequest`
|
||||
- `AdvancedNarrationResult`
|
||||
- `ServiceHealth`
|
||||
- `NarrationRequest`
|
||||
- `SynthesisResult`
|
||||
- `SessionConfig`
|
||||
- `TranscriptMetadata`
|
||||
- `TranscriptResult`
|
||||
- existing `CubeDesk` clean-room regression suite
|
||||
|
||||
## Queue effect
|
||||
|
||||
This packet consumes the current `Phase 6R-H` implementation slice.
|
||||
|
||||
`coqui-ai/TTS` remains only partially incorporated:
|
||||
|
||||
- landed now:
|
||||
- advanced narration orchestration profile
|
||||
- normalized model/vocoder binding metadata
|
||||
- multilingual routing and speaker-selection flags
|
||||
- reference-speaker conditioning flag
|
||||
- bounded mock and HTTP voice-client seam support for the richer narration profile
|
||||
- advanced voice service-health capability exposure
|
||||
- still deferred:
|
||||
- downloadable voice/model review
|
||||
- bundled voice asset shipping
|
||||
- voice conversion
|
||||
- broad playback runtime ownership
|
||||
- viseme / gesture runtime integration
|
||||
- broad assistant-platform scope
|
||||
|
||||
The next clean move is:
|
||||
|
||||
- source-backed `Phase 6R-I` `roice3/Magic120Cell` preparation/control pass
|
||||
|
||||
Keep the future sequencing guards visible:
|
||||
|
||||
- speech-input / voice sidecar set
|
||||
- keep code-license judgments separate from model, voice, and payload-license review
|
||||
- `coqui-ai/TTS`
|
||||
- keep the landed slice bounded to orchestration metadata rather than broad voice-platform
|
||||
ownership
|
||||
- `roice3/Magic120Cell`
|
||||
- keep the next widening slice bounded to one retained hyper-engine family rather than treating
|
||||
the row as a generic non-Euclidean platform by default
|
||||
|
|
@ -0,0 +1,79 @@
|
|||
# HyperTwist Phase 6R-H coqui-ai/TTS advanced narration preparation packet
|
||||
|
||||
Created on `2026-05-22`
|
||||
|
||||
## Status
|
||||
|
||||
- first-party HyperTwist packet
|
||||
- source-backed permissive `Phase 6R-H` preparation/control slice
|
||||
|
||||
## Purpose
|
||||
|
||||
This packet narrows the retained `coqui-ai/TTS` row before implementation.
|
||||
|
||||
The retained first slice is:
|
||||
|
||||
- advanced narration orchestration profile above the already landed embodied companion narration lane
|
||||
|
||||
It is not:
|
||||
|
||||
- a broad TTS platform packet
|
||||
- a downloadable voice-asset or model-shipping packet
|
||||
- a provider-profile or BYOK packet
|
||||
- a voice-conversion packet
|
||||
- a viseme, gesture, or avatar-runtime widening packet
|
||||
|
||||
## Source-backed basis
|
||||
|
||||
Primary donor surfaces inspected:
|
||||
|
||||
- `README.md`
|
||||
- `docs/source/inference.md`
|
||||
- `TTS/server/README.md`
|
||||
- `TTS/api.py`
|
||||
|
||||
## Narrowing decision
|
||||
|
||||
The strongest bounded donor seam is not raw audio output by itself. That lighter local sidecar lane
|
||||
already landed through `rhasspy/piper`.
|
||||
|
||||
The smallest stronger `coqui-ai/TTS` seam is:
|
||||
|
||||
- richer narration orchestration metadata
|
||||
- normalized model/vocoder binding selection
|
||||
- speaker-selection and multilingual routing flags
|
||||
- optional speaker-reference conditioning
|
||||
- sentence-splitting policy above the same first-party narration contract
|
||||
|
||||
The first bounded implementation packet should therefore land only:
|
||||
|
||||
- first-party advanced narration orchestration profile types
|
||||
- bounded request/result metadata for model/vocoder/speaker/reference selection
|
||||
- mock and HTTP voice-client seam support for the bounded advanced narration profile
|
||||
- focused automation around the advanced narration request/result/health contract
|
||||
|
||||
## Explicit deferrals
|
||||
|
||||
Still deferred after this control pass:
|
||||
|
||||
- downloadable voice/model review
|
||||
- bundled voice asset shipping
|
||||
- voice conversion
|
||||
- broad playback runtime ownership
|
||||
- viseme / gesture runtime integration
|
||||
- provider-profile and BYOK custody
|
||||
- broad assistant-platform scope
|
||||
|
||||
## Queue effect
|
||||
|
||||
This preparation packet authorizes one bounded permissive implementation packet:
|
||||
|
||||
- `Phase 6R-H` advanced narration orchestration profile
|
||||
|
||||
After that packet, move the queue head to:
|
||||
|
||||
- source-backed `Phase 6R-I` `roice3/Magic120Cell` preparation/control pass
|
||||
|
||||
Keep the legal guard visible:
|
||||
|
||||
- `coqui-ai/TTS` code-license judgment stays separate from model, voice, and payload review
|
||||
|
|
@ -80,10 +80,10 @@ Use these as the current governing docs:
|
|||
|
||||
The next bounded move is now:
|
||||
|
||||
1. source-backed `Phase 6R-H` `coqui-ai/TTS` preparation/control pass
|
||||
2. keep it behind the landed `whisper.cpp`, `faster-whisper`, and `piper` bounded slices
|
||||
3. narrow the first widening slice to one richer bounded voice-output seam
|
||||
4. keep voice-asset and payload review separate from the code-license judgment
|
||||
1. source-backed `Phase 6R-I` `roice3/Magic120Cell` preparation/control pass
|
||||
2. keep it after the landed `whisper.cpp`, `faster-whisper`, `piper`, and `coqui-ai/TTS` bounded slices
|
||||
3. narrow the first widening slice to one bounded hyper-engine topology/runtime-anchor seam
|
||||
4. keep later non-Euclidean shell widening separate from the code-license judgment and retained row ownership
|
||||
|
||||
## Memory-specific sequencing rule
|
||||
|
||||
|
|
|
|||
|
|
@ -183,6 +183,7 @@ repo.
|
|||
| Provider-backed speech session health and transcript envelopes | Implemented now | first-party speech client surfaces | Current first-party speech seam. |
|
||||
| Python transcription-service orchestration profile | Implemented now | landed `faster-whisper` packet | Current bounded Python STT orchestration truth above the existing speech session boundary. |
|
||||
| Local narration sidecar seam | Implemented now | landed `piper` packet | Current bounded offline narration contract, voice catalog, and local sidecar truth. |
|
||||
| Advanced narration orchestration profile | Implemented now | landed `coqui-ai/TTS` packet | Current bounded richer narration profile above the existing companion narration contract. |
|
||||
| Broad voice output / narration sidecars | Deep-source grounded retained | `piper`, `coqui-ai/TTS` retained remainder | Keep downloadable voice/model review separate from the code-license judgment. |
|
||||
| Broad live microphone shell and downloadable payload shipping | Deep-source grounded retained | speech stack retained remainder | Explicitly deferred. |
|
||||
|
||||
|
|
|
|||
|
|
@ -78,8 +78,13 @@ Canonical discovery surfaces for roadmap interpretation:
|
|||
in current code
|
||||
- `rhasspy/piper` remains partially incorporated; downloadable voice/model review, bundled voice
|
||||
asset shipping, and broad voice-platform ownership stay deferred
|
||||
- the current next bounded move is a source-backed `Phase 6R-H` `coqui-ai/TTS`
|
||||
preparation/control pass for a richer bounded voice-output seam
|
||||
- the generic source-backed `Phase 6R-H` `coqui-ai/TTS` control pass is now consumed
|
||||
- the bounded permissive `Phase 6R-H` `coqui-ai/TTS` advanced narration orchestration packet is now
|
||||
landed in current code
|
||||
- `coqui-ai/TTS` remains partially incorporated; downloadable voice/model review, bundled voice
|
||||
asset shipping, voice conversion, and broad voice-platform ownership stay deferred
|
||||
- the current next bounded move is a source-backed `Phase 6R-I` `roice3/Magic120Cell`
|
||||
preparation/control pass for one bounded hyper-engine widening slice
|
||||
- the repo-row implementation queue is now live from that `Phase 6R-A` entry point rather than
|
||||
waiting on another first-party packet
|
||||
|
||||
|
|
@ -141,8 +146,8 @@ Current routing truth:
|
|||
- active non-live implementation-board rows: `35`
|
||||
- retained benchmark, oracle, or clean-room-later rows outside the active implementation board: `9`
|
||||
|
||||
The next bounded move is a source-backed `Phase 6R-H`
|
||||
`coqui-ai/TTS` preparation/control pass.
|
||||
The next bounded move is a source-backed `Phase 6R-I`
|
||||
`roice3/Magic120Cell` preparation/control pass.
|
||||
|
||||
Queue interpretation after that packet:
|
||||
|
||||
|
|
@ -225,13 +230,32 @@ Queue interpretation after that packet:
|
|||
- rich playback runtime ownership
|
||||
- viseme / gesture runtime integration
|
||||
- broad assistant-platform scope
|
||||
- `coqui-ai/TTS` is now landed as a partially incorporated row rather than a generic future
|
||||
placeholder:
|
||||
- landed:
|
||||
- advanced narration orchestration profile
|
||||
- normalized model/vocoder binding metadata
|
||||
- multilingual routing and speaker-selection flags
|
||||
- reference-speaker conditioning flag
|
||||
- bounded mock and HTTP voice-client seam support for the richer narration profile
|
||||
- advanced voice service-health capability exposure
|
||||
- still deferred:
|
||||
- downloadable voice/model review
|
||||
- bundled voice asset shipping
|
||||
- voice conversion
|
||||
- broad playback runtime ownership
|
||||
- viseme / gesture runtime integration
|
||||
- broad assistant-platform scope
|
||||
- the next queue shape is now:
|
||||
- `coqui-ai/TTS`
|
||||
- `roice3/Magic120Cell`
|
||||
- keep the speech-input / voice sidecar legal sequencing guard visible:
|
||||
- keep code-license judgments separate from model, voice, and payload-license review
|
||||
- keep the provider-neutral speech-lane guard visible:
|
||||
- first-party HyperTwist owns the top-level provider/session contract
|
||||
- `whisper.cpp` and `faster-whisper` may win narrower donor slices without inheriting that lane
|
||||
- keep the voice-output guard visible:
|
||||
- `coqui-ai/TTS` stays bounded to advanced narration orchestration metadata rather than broad
|
||||
voice-platform ownership
|
||||
- keep the optional `Phase 3R-G` browser comparison lane deferred unless the landed primary browser
|
||||
spatial stack exposes a real gap
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue