core_c_api.h

Go to the documentation of this file.

/* Copyright 2021 ID R&D Inc. All Rights Reserved. */

#pragma once

#include <stdint.h>
#include <stdbool.h>
#include <stdlib.h>

#include "config.h"

#ifdef __cplusplus
extern "C" {
#endif

typedef enum VoiceSdkErrorCode {
    kVoiceSdkErrorCodeOk = 0, 
    kVoiceSdkErrorCodeNullPointer = 1, 
    kVoiceSdkErrorCodeInvalidArgument = 2, 
    kVoiceSdkErrorCodeRuntimeError = 3, 
    kVoiceSdkErrorCodeLicenseError = 4 
} VoiceSdkErrorCode;

#define DECL_TYPED_ARRAY(name, type)                     \
    typedef struct {                                     \
        size_t size;                                     \
        type* data;                                      \
    } name;                                              \
                                                         \
    VOICE_SDK_API                                        \
    name* name##Create(size_t size);                     \
                                                         \
    VOICE_SDK_API                                        \
    void name##Release(name* wrapped);                   \
                                                         \
    /* nested create/release are only used internally */ \
    void name##CreateNested(name* wrapped, size_t size); \
    void name##ReleaseNested(name* wrapped);             \
                                                         \

DECL_TYPED_ARRAY(VoiceSdkByteArray, uint8_t)
DECL_TYPED_ARRAY(VoiceSdkShortArray, int16_t)
DECL_TYPED_ARRAY(VoiceSdkFloatArray, float)

typedef struct VoiceSdkVoiceTemplate VoiceSdkVoiceTemplate;

typedef struct VoiceSdkBuildInfo {
    const char* version;

    const char* components;

    const char* git_info;
} VoiceSdkBuildInfo;

typedef enum VoiceSdkLicenseFeature {
    kVoiceSdkLicenseFeatureCore = 7,         
    kVoiceSdkLicenseFeatureVerification = 2, 
    kVoiceSdkLicenseFeatureLivenessPresentationAttackDetection =
            4, 
    kVoiceSdkLicenseFeatureLivenessVoiceClonesDetection = 31, 
    kVoiceSdkLicenseFeatureQualityChecking = 6 
} VoiceSdkLicenseFeature;

typedef struct VoiceSdkLicenseFeatureInfo {
    VoiceSdkLicenseFeature feature; 
    char expiration_date[11];       
} VoiceSdkLicenseFeatureInfo;

typedef struct VoiceSdkLicenseExpirationDate {
    char license_expiration_date[11];
} VoiceSdkLicenseExpirationDate;

typedef struct VoiceSdkAudioInterval {
    size_t start_sample;

    size_t end_sample;

    size_t start_time;

    size_t end_time;

    int sample_rate;
} VoiceSdkAudioInterval;

typedef struct VoiceSdkAudioInfo {
    size_t num_samples;

    size_t sample_rate;

    size_t num_channels;
} VoiceSdkAudioInfo;

typedef enum VoiceSdkChannelType {
    CHANNEL_TYPE_MIC = 1,
    CHANNEL_TYPE_TEL = 2,
    CHANNEL_TYPE_MIXED = 3
} VoiceSdkChannelType;

/********************
 * Build Info calls *
 ********************/

VOICE_SDK_API
void VoiceSdkGetBuildInfo(VoiceSdkBuildInfo* build_info);

/******************
 * Settings calls *
 ******************/

VOICE_SDK_API
void VoiceSdkSetNumThreads(unsigned int num_threads);

VOICE_SDK_API
void VoiceSdkSetAllowMobileGpu(bool allow_mobile_gpu);

VOICE_SDK_API
void VoiceSdkSetAllowMobileNnapi(bool allow_mobile_nnapi);

VOICE_SDK_API
void VoiceSdkSetUseVoiceTemplateCompression(bool use_voice_template_compression);

VOICE_SDK_API
bool VoiceSdkSetMobileLicense(const char* mobile_license_string, char** error_msg);

VOICE_SDK_API
VoiceSdkLicenseFeatureInfo* VoiceSdkGetLicenseInfo(size_t* num_license_features, VoiceSdkErrorCode* error_code,
                                                   char** error_msg);

VOICE_SDK_API
void VoiceSdkLicenseFeatureInfoArrayRelease(VoiceSdkLicenseFeatureInfo* license_info);

VOICE_SDK_API
VoiceSdkLicenseExpirationDate VoiceSdkGetLicenseExpirationDate();

/*******************
 * Wav Utils calls *
 *******************/

VOICE_SDK_API
VoiceSdkByteArray* VoiceSdkWavUtilsReadFromFileAsPcm16Bytes(const char* wav_path, size_t* sample_rate,
                                                            char** error_msg);

VOICE_SDK_API
VoiceSdkShortArray* VoiceSdkWavUtilsReadFromFileAsPcm16Samples(const char* wav_path, size_t* sample_rate,
                                                               char** error_msg);

VOICE_SDK_API
VoiceSdkFloatArray* VoiceSdkWavUtilsReadFromFileAsFloatSamples(const char* wav_path, size_t* sample_rate,
                                                               char** error_msg);

VOICE_SDK_API
VoiceSdkFloatArray* VoiceSdkWavUtilsReadFromFileAs16BitFloatSamples(const char* wav_path, size_t* sample_rate,
                                                                    char** error_msg);

VOICE_SDK_API
VoiceSdkByteArray* VoiceSdkWavUtilsReadFromBufferAsPcm16Bytes(const uint8_t* data, size_t data_size,
                                                              size_t* sample_rate, char** error_msg);

VOICE_SDK_API
VoiceSdkShortArray* VoiceSdkWavUtilsReadFromBufferAsPcm16Samples(const uint8_t* data, size_t data_size,
                                                                 size_t* sample_rate, char** error_msg);

VOICE_SDK_API
VoiceSdkFloatArray* VoiceSdkWavUtilsReadFromBufferAsFloatSamples(const uint8_t* data, size_t data_size,
                                                                 size_t* sample_rate, char** error_msg);

VOICE_SDK_API
VoiceSdkFloatArray* VoiceSdkWavUtilsReadFromBufferAs16BitFloatSamples(const uint8_t* data, size_t data_size,
                                                                      size_t* sample_rate, char** error_msg);
VOICE_SDK_API
bool VoiceSdkWavUtilsGetAudioInfoFromFile(const char* wav_path, VoiceSdkAudioInfo* result, char** error_msg);

VOICE_SDK_API
bool VoiceSdkWavUtilsGetAudioInfoFromBuffer(const uint8_t* data, size_t data_size, VoiceSdkAudioInfo* result,
                                            char** error_msg);

/************************
 * Voice Template calls *
 ************************/

VOICE_SDK_API
VoiceSdkVoiceTemplate* VoiceSdkVoiceTemplateLoadFromFile(const char* filepath, char** error_msg);

VOICE_SDK_API
bool VoiceSdkVoiceTemplateSaveToFile(const VoiceSdkVoiceTemplate* voice_template, const char* filepath,
                                     char** error_msg);

VOICE_SDK_API
VoiceSdkVoiceTemplate* VoiceSdkVoiceTemplateDeserialize(const uint8_t* array, size_t array_size, char** error_msg);

VOICE_SDK_API
VoiceSdkByteArray* VoiceSdkVoiceTemplateSerialize(const VoiceSdkVoiceTemplate* voice_template, char** error_msg);

VOICE_SDK_API
bool VoiceSdkVoiceTemplateGetChannelType(const VoiceSdkVoiceTemplate* voice_template, VoiceSdkChannelType* channel_type,
                                         char** error_msg);

VOICE_SDK_API
const char* VoiceSdkVoiceTemplateGetInitDataId(const VoiceSdkVoiceTemplate* voice_template, char** error_msg);

VOICE_SDK_API
bool VoiceSdkVoiceTemplateIsValid(const VoiceSdkVoiceTemplate* voice_template, bool* is_valid, char** error_msg);

VOICE_SDK_API
void VoiceSdkVoiceTemplateRelease(VoiceSdkVoiceTemplate* voice_template);

VOICE_SDK_API
VoiceSdkShortArray* VoiceSdkOpusUtilsReadFromBufferAsPcm16Samples(const uint8_t* data, size_t data_size,
                                                                  size_t* sample_rate, char** error_msg);

#ifdef __cplusplus
}
#endif