AudioTypes.h

// OpenAI, Copyright LifeEXE. All Rights Reserved.
 
#pragma once
 
#include "CoreMinimal.h"
#include "AudioTypes.generated.h"
 
UENUM(BlueprintType)
enum class EAudioModel : uint8
{
    Whisper_1
};
 
UENUM(BlueprintType)
enum class ETTSModel : uint8
{
    TTS_1,
    TTS_1_HD,
    GPT_4O_MINI_TTS,
    GPT_4O_MINI_TTS_2025_12_15
};
 
UENUM(BlueprintType)
enum class EVoice : uint8
{
    Alloy,
    Ash,
    Ballad,
    Cedar,
    Coral,
    Echo,
    Fable,
    Marin,
    Nova,
    Onyx,
    Sage,
    Shimmer,
    Verse,
};
 
// Make sure that WMFCodecs plugin is enabled (Edit->Plugins->WMFCodecs)
UENUM(BlueprintType)
enum class ETTSAudioFormat : uint8
{
    MP3,
    OPUS,  // Not supported by UnrealEngine
    AAC,
    FLAC,  // Not supported by UnrealEngine
    WAV,
    PCM  // Not supported by UnrealEngine
};
 
UENUM(BlueprintType)
enum class ETranscriptFormat : uint8
{
    JSON,
    Text,
    Str,
    Verbose_JSON,
    Vtt
};

//                 REQUEST TYPES
 
USTRUCT(BlueprintType)
struct FAudioBase
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString File;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Model{"whisper-1"};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Prompt;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Response_Format{"json"};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    float Temperature{0.0f};
};
 
USTRUCT(BlueprintType)
struct FAudioTranscription : public FAudioBase
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Language;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    TArray<int32> Timestamp_Granularities;
};
 
USTRUCT(BlueprintType)
struct FAudioTranslation : public FAudioBase
{
    GENERATED_BODY()
};
 
USTRUCT(BlueprintType)
struct FSpeech
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Model{"tts-1"};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Input;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Voice{"alloy"};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Instructions{};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Response_Format{"mp3"};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    float Speed{1.0f};

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Optional")
    FString Stream_Format{};
};
 
USTRUCT(BlueprintType)
struct FCreateVoice
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Audio_Sample;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Consent;

    UPROPERTY(BlueprintReadWrite, Category = "OpenAI | Required")
    FString Name;
};

//                 RESPONSE TYPES

USTRUCT(BlueprintType)
struct FAudioTranscriptionResponse
{
    GENERATED_BODY()
 
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Text;
};
 
USTRUCT(BlueprintType)
struct FAudioTranscriptionWord
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Word;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    float Start{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    float End{};
};
 
USTRUCT(BlueprintType)
struct FAudioTranscriptionSegment
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    int32 Id{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    int32 Seek{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    float Start{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    float End{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Text;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    TArray<int32> Tokens;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    double Temperature{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    double Avg_Logprob{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    double Compression_Ratio{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    double No_Speech_Prob{};
};

USTRUCT(BlueprintType)
struct FAudioTranscriptionVerboseResponse
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Language;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    float Duration{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Text;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    TArray<FAudioTranscriptionWord> Words;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    TArray<FAudioTranscriptionSegment> Segments;
};
 
USTRUCT(BlueprintType)
struct FAudioTranslationResponse
{
    GENERATED_BODY()
 
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Text;
};
 
USTRUCT(BlueprintType)
struct FSpeechResponse
{
    GENERATED_BODY()
 
    UPROPERTY(BlueprintReadWrite, Category = "OpenAI")
    TArray<uint8> Bytes;
};
 
USTRUCT(BlueprintType)
struct FCreateVoiceResponse
{
    GENERATED_BODY()
 
    
    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Id;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    int32 Created_At{};

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Name;

    UPROPERTY(BlueprintReadOnly, Category = "OpenAI")
    FString Object;
};