Espeak SAPI/dll 在 Windows 上的使用？

Question

问题：我正在尝试使用 espeak 文本转语音引擎。 所以我让它在 linux 上工作得很好（下面的代码）。 现在我也想把这个基本程序移植到 Windows 上，但这几乎是不可能的......

部分问题是windows dll只允许AUDIO_OUTPUT_SYNCHRONOUS，这意味着它需要回调，但我不知道如何从回调中播放音频......首先它崩溃了，然后我意识到，我需要一个回调函数，现在我在回调函数中获取数据，但我不知道如何播放它...因为它既不是 wav 文件，也不是像 Linux 上那样自动播放。

sourceforge 站点比较没用，因为它基本上说使用 SAPI 版本，但是没有关于如何使用 sapi espeak dll 的示例...

无论如何，这是我的代码，有人可以帮忙吗？

#ifdef __cplusplus
#include <cstdio>
#include <cstdlib>
#include <cstring>
#else
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#endif

#include <assert.h>
#include <ctype.h>

//#include "speak_lib.h"
#include "espeak/speak_lib.h"

// libespeak-dev: /usr/include/espeak/speak_lib.h
// apt-get install libespeak-dev
// apt-get install libportaudio-dev

// g++ -o mine mine.cpp -lespeak
// g++ -o mine mine.cpp -I/usr/include/espeak/ -lespeak
// gcc -o mine mine.cpp -I/usr/include/espeak/ -lespeak


char voicename[40];
int samplerate;
int quiet = 0;
static char genders[4] = {' ','M','F',' '};

//const char *data_path = "/usr/share/";   // /usr/share/espeak-data/
const char *data_path = NULL;   // use default path for espeak-data


int strrcmp(const char *s, const char *sub)
{
int slen = strlen(s);
int sublen = strlen(sub);
return memcmp(s + slen - sublen, sub, sublen);
}


char * strrcpy(char *dest, const char *source)
{
// Pre assertions
assert(dest != NULL);
assert(source != NULL);
assert(dest != source);

// tk: parentheses
while((*dest++ = *source++))
    ;
return(--dest);
}

const char* GetLanguageVoiceName(const char* pszShortSign)
{
#define LANGUAGE_LENGTH 30
static char szReturnValue[LANGUAGE_LENGTH] ;
memset(szReturnValue, 0, LANGUAGE_LENGTH);

for (int i = 0; pszShortSign[i] != '\0'; ++i)
    szReturnValue[i] = (char) tolower(pszShortSign[i]);

const espeak_VOICE **voices;
espeak_VOICE voice_select;
voices = espeak_ListVoices(NULL);

const espeak_VOICE *v;
for(int ix=0; (v = voices[ix]) != NULL; ix++)
{
    if( !strrcmp( v->languages, szReturnValue) )
    {
        strcpy(szReturnValue, v->name);
        return szReturnValue;
    }
} // End for

strcpy(szReturnValue, "default");
return szReturnValue;
} // End function getvoicename


void ListVoices()
{
const espeak_VOICE **voices;
espeak_VOICE voice_select;
voices = espeak_ListVoices(NULL);

const espeak_VOICE *v;
for(int ix=0; (v = voices[ix]) != NULL; ix++)
{
    printf("Shortsign: %s\n", v->languages);
    printf("age: %d\n", v->age);
    printf("gender: %c\n", genders[v->gender]);
    printf("name: %s\n", v->name);
    printf("\n\n");
} // End for
} // End function getvoicename


int main()
{
printf("Hello World!\n");
const char* szVersionInfo = espeak_Info(NULL);

printf("Espeak version: %s\n", szVersionInfo);
samplerate = espeak_Initialize(AUDIO_OUTPUT_PLAYBACK,0,data_path,0);

strcpy(voicename, "default");
// espeak --voices
strcpy(voicename, "german");
strcpy(voicename, GetLanguageVoiceName("DE"));

if(espeak_SetVoiceByName(voicename) != EE_OK)
{
    printf("Espeak setvoice error...\n");
}

static char word[200] = "Hello World" ;
strcpy(word, "TV-fäns aufgepasst, es ist 20 Uhr 15. Zeit für Rambo 3");
strcpy(word, "Unnamed Player wurde zum Opfer von GSG9");
int speed = 220;
int volume = 500; // volume in range 0-100    0=silence
int pitch = 50; //  base pitch, range 0-100.  50=normal

// espeak.cpp 625
espeak_SetParameter(espeakRATE, speed, 0);
espeak_SetParameter(espeakVOLUME,volume,0);
espeak_SetParameter(espeakPITCH,pitch,0);
// espeakRANGE:   pitch range, range 0-100. 0-monotone, 50=normal
// espeakPUNCTUATION:  which punctuation characters to announce:
    // value in espeak_PUNCT_TYPE (none, all, some), 
espeak_VOICE *voice_spec = espeak_GetCurrentVoice();
voice_spec->gender=2; // 0=none 1=male, 2=female,
//voice_spec->age = age;

espeak_SetVoiceByProperties(voice_spec);


espeak_Synth( (char*) word, strlen(word)+1, 0, POS_CHARACTER, 0, espeakCHARS_AUTO, NULL, NULL);
espeak_Synchronize();

strcpy(voicename, GetLanguageVoiceName("EN"));
espeak_SetVoiceByName(voicename);
strcpy(word, "Geany was fragged by GSG9 Googlebot");
strcpy(word, "Googlebot");

espeak_Synth( (char*) word, strlen(word)+1, 0, POS_CHARACTER, 0, espeakCHARS_AUTO, NULL, NULL);
espeak_Synchronize();


espeak_Terminate();
printf("Espeak terminated\n");
return EXIT_SUCCESS; 
}

/*
if(espeak_SetVoiceByName(voicename) != EE_OK)
{
    memset(&voice_select,0,sizeof(voice_select));
    voice_select.languages = voicename;
    if(espeak_SetVoiceByProperties(&voice_select) != EE_OK)
    {
        fprintf(stderr,"%svoice '%s'\n",err_load,voicename);
        exit(2);
    }
}
*/

上面的代码适用于 Linux。 下面的代码是关于我在 Vista x64（32 位 emu）上的代码：

#ifdef __cplusplus
#include <cstdio>
#include <cstdlib>
#include <cstring>
#else
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#endif

#include <assert.h>
#include <ctype.h>

#include "speak_lib.h"
//#include "espeak/speak_lib.h"

// libespeak-dev: /usr/include/espeak/speak_lib.h
// apt-get install libespeak-dev
// apt-get install libportaudio-dev

// g++ -o mine mine.cpp -lespeak
// g++ -o mine mine.cpp -I/usr/include/espeak/ -lespeak
// gcc -o mine mine.cpp -I/usr/include/espeak/ -lespeak


char voicename[40];
int iSampleRate;
int quiet = 0;
static char genders[4] = {' ','M','F',' '};

//const char *data_path = "/usr/share/";   // /usr/share/espeak-data/
//const char *data_path = NULL;   // use default path for espeak-data
const char *data_path = "C:\\Users\\Username\\Desktop\\espeak-1.43-source\\espeak-1.43-source\\";


int strrcmp(const char *s, const char *sub)
{
int slen = strlen(s);
int sublen = strlen(sub);
return memcmp(s + slen - sublen, sub, sublen);
}


char * strrcpy(char *dest, const char *source)
{
// Pre assertions
assert(dest != NULL);
assert(source != NULL);
assert(dest != source);

// tk: parentheses
while((*dest++ = *source++))
    ;
return(--dest);
}

const char* GetLanguageVoiceName(const char* pszShortSign)
{
#define LANGUAGE_LENGTH 30
static char szReturnValue[LANGUAGE_LENGTH] ;
memset(szReturnValue, 0, LANGUAGE_LENGTH);

for (int i = 0; pszShortSign[i] != '\0'; ++i)
    szReturnValue[i] = (char) tolower(pszShortSign[i]);

const espeak_VOICE **voices;
espeak_VOICE voice_select;
voices = espeak_ListVoices(NULL);

const espeak_VOICE *v;
for(int ix=0; (v = voices[ix]) != NULL; ix++)
{
    if( !strrcmp( v->languages, szReturnValue) )
    {
        strcpy(szReturnValue, v->name);
        return szReturnValue;
    }
} // End for

strcpy(szReturnValue, "default");
return szReturnValue;
} // End function getvoicename


void ListVoices()
{
const espeak_VOICE **voices;
espeak_VOICE voice_select;
voices = espeak_ListVoices(NULL);

const espeak_VOICE *v;
for(int ix=0; (v = voices[ix]) != NULL; ix++)
{
    printf("Shortsign: %s\n", v->languages);
    printf("age: %d\n", v->age);
    printf("gender: %c\n", genders[v->gender]);
    printf("name: %s\n", v->name);
    printf("\n\n");
} // End for
} // End function getvoicename


/* Callback from espeak.  Directly speaks using AudioTrack. */
#define LOGI(x) printf("%s\n", x)
static int AndroidEspeakDirectSpeechCallback(short *wav, int numsamples, espeak_EVENT *events) 
{
    char buf[100];
    sprintf(buf, "AndroidEspeakDirectSpeechCallback: %d samples", numsamples);
    LOGI(buf);

    if (wav == NULL) 
{
        LOGI("Null: speech has completed");
    }

    if (numsamples > 0)
{
        //audout->write(wav, sizeof(short) * numsamples);
        sprintf(buf, "AudioTrack wrote: %d bytes", sizeof(short) * numsamples);
        LOGI(buf);
    }

    return 0;  // continue synthesis (1 is to abort)
}


static int AndroidEspeakSynthToFileCallback(short *wav, int numsamples,espeak_EVENT *events) 
{
    char buf[100];
    sprintf(buf, "AndroidEspeakSynthToFileCallback: %d samples", numsamples);
    LOGI(buf);

    if (wav == NULL) 
{
        LOGI("Null: speech has completed");
    }

    // The user data should contain the file pointer of the file to write to
    //void* user_data = events->user_data;
FILE* user_data = fopen ( "myfile1.wav" , "ab" );

    FILE* fp = static_cast<FILE *>(user_data);

    // Write all of the samples
    fwrite(wav, sizeof(short), numsamples, fp);
    return 0;  // continue synthesis (1 is to abort)
}



int main()
{
printf("Hello World!\n");
const char* szVersionInfo = espeak_Info(NULL);

printf("Espeak version: %s\n", szVersionInfo);

iSampleRate = espeak_Initialize(AUDIO_OUTPUT_SYNCHRONOUS, 4096, data_path, 0);
if (iSampleRate <= 0) 
{
    printf("Unable to initialize espeak");
    return EXIT_FAILURE;
}

//samplerate = espeak_Initialize(AUDIO_OUTPUT_PLAYBACK,0,data_path,0);

//ListVoices();

strcpy(voicename, "default");
// espeak --voices
//strcpy(voicename, "german");
//strcpy(voicename, GetLanguageVoiceName("DE"));

if(espeak_SetVoiceByName(voicename) != EE_OK)
{
    printf("Espeak setvoice error...\n");
}

static char word[200] = "Hello World" ;
strcpy(word, "TV-fäns aufgepasst, es ist 20 Uhr 15. Zeit für Rambo 3");
strcpy(word, "Unnamed Player wurde zum Opfer von GSG9");
int speed = 220;
int volume = 500; // volume in range 0-100    0=silence
int pitch = 50; //  base pitch, range 0-100.  50=normal


// espeak.cpp 625
espeak_SetParameter(espeakRATE, speed, 0);
espeak_SetParameter(espeakVOLUME,volume,0);
espeak_SetParameter(espeakPITCH,pitch,0);
// espeakRANGE:   pitch range, range 0-100. 0-monotone, 50=normal
// espeakPUNCTUATION:  which punctuation characters to announce:
    // value in espeak_PUNCT_TYPE (none, all, some), 
//espeak_VOICE *voice_spec = espeak_GetCurrentVoice();
//voice_spec->gender=2; // 0=none 1=male, 2=female,
//voice_spec->age = age;

//espeak_SetVoiceByProperties(voice_spec);

//espeak_SetSynthCallback(AndroidEspeakDirectSpeechCallback);
espeak_SetSynthCallback(AndroidEspeakSynthToFileCallback);

unsigned int unique_identifier;
espeak_ERROR err = espeak_Synth( (char*) word, strlen(word)+1, 0, POS_CHARACTER, 0, espeakCHARS_AUTO, &unique_identifier, NULL);

err = espeak_Synchronize();



/*
strcpy(voicename, GetLanguageVoiceName("EN"));
espeak_SetVoiceByName(voicename);
strcpy(word, "Geany was fragged by GSG9 Googlebot");
strcpy(word, "Googlebot");

espeak_Synth( (char*) word, strlen(word)+1, 0, POS_CHARACTER, 0, espeakCHARS_AUTO, NULL, NULL);
espeak_Synchronize();
*/

// espeak_Cancel();
espeak_Terminate();
printf("Espeak terminated\n");
system("pause");
return EXIT_SUCCESS; 
}

Answer 1

需要对源代码进行几处更改才能使 windows 库具有与 Linux 上的库相同的功能。我列出了更改here。也可以使用即用型二进制文件。

所有补丁和描述也发送给了espeak维护者（通过邮件列表和补丁跟踪器公开），所以将来可能会直接可用。

Answer 2

您是否尝试过将回调中获得的缓冲区传递给sndplaysnd()？？

Declare Function sndPlaySound Lib "winmm.dll" Alias "sndPlaySoundA" (ByVal lpszSoundName As String, ByVal uFlags As Long) As Long

其标准winAPI如下：

  sndPlaySound(buffer[0], SND_ASYNC | SND_MEMORY)

或者，如果您有一个可以播放音频的 wav 文件：

  sndPlaySound(filename, SND_ASYNC)

playsound 具有 ASYNC 模式，在播放音频时不会阻止程序的执行。

注意：我在VB中使用过，上面的sn-ps是在VB中使用的。如果您使用 VC++ 进行编码，则可能必须相应地修改它们。但基本意图保持不变；将缓冲区传递给设置了 ASYNC 标志的 sndPlaySound。

祝你好运！！