text-to-speech на Паскале

C300G — Tue, 29 Jun 2004 13:38:08 +0000

C300G: Если ещё нужно (SAPI под Dlphi):
Brian Long "Speech Synthesis & Speech Recognition".
Там в разделе "SAPI 4" даны ссылки на "адаптированные" под Паскаль speech.h и spchtel.h.

off: Если постишь исходный код, используй тэги CODE.

text-to-speech на Паскале

Гость Leo — Thu, 29 Apr 2004 12:09:38 +0000

Гость Leo: Кто-нибудь адаптировал Speech.h под Паскаль? Вроде, все содрал из SDK, работает до Select, а потом облом. C-шный исходник, с который использовал, прекрасно компилится и работает, так что дело не в SAPI или железе. Может, кто опытный свежим взглядом что увидит?

PROGRAM Text_To_Speech_Test;

USES Messages,Windows,ActiveX,Strings,MMSystem,WinServ;

{$A+}
{$D-}
CONST
SEVERITY=$80000000;
TTSI_NAMELEN=262;
TTSI_STYLELEN=262;
{SAPI class indentifiers}
CLSID_TTSEnumerator: TGUID = '{d67c0280-c743-11cd-80e5-00aa003e4b50}';
CLSID_MMAudioDest: TGUID = '{cb96b400-c743-11cd-80e5-00aa003e4b50}';
IID_IAudioMultiMediaDevice: TGUID = '{b68ad320-c743-11cd-80e5-00aa003e4b50}';
IID_ITTSEnum: TGUID = '{05EB6C6D-DBAB-11CD-B3CA-00AA0047BA4F}';
IID_ITTSCentral: TGUID = '{05EB6C6A-DBAB-11CD-B3CA-00AA0047BA4F}';

TYPE
pUnknown=^iUnknown;
SDATA = record
pData: pointer;
dwSize: DWORD;
end;
TLANGUAGE = record
LanguageID: LangID;
szDialect: array[0..63] of AnsiChar;
end;
VOICECHARSET = UINT;
QWORD = Int64;

TTSMODEINFO = record
gEngineID : TGUID;
szMfgName : array [0..TTSI_NAMELEN-1] of AnsiChar;
szProductName : array [0..TTSI_NAMELEN-1] of AnsiChar;
gModeID : TGUID;
szModeName : array [0..TTSI_NAMELEN-1] of AnsiChar;
Language : TLanguage;
szSpeaker : array [0..TTSI_NAMELEN-1] of AnsiChar;
szStyle : array [0..TTSI_STYLELEN-1] of AnsiChar;
wGender : WORD;
wAge : WORD;
dwFeatures : DWORD;
dwInterfaces : DWORD;
dwEngineFeatures : DWORD;
end;

PTTSMODEINFO = ^TTSMODEINFO;
{Audio interface}
IAudioMultiMediaDevice = interface(IUnknown)
['{B68AD320-C743-11cd-80E5-00AA003E4B50}']
function CustomMessage(uMsg: UINT; dData: SDATA): HResult; stdcall;
function DeviceNumGet(var dwDeviceID: DWORD): HResult; stdcall;
function DeviceNumSet(dwDeviceID: DWORD): HResult; stdcall;
end;
{Central interface}
ITTSCentral = interface(IUnknown)
['{05EB6C6A-DBAB-11CD-B3CA-00AA0047BA4F}']
function Inject(pszTag: PAnsiChar): HResult; stdcall;
function ModeGet(var ttsInfo: TTSModeInfo): HResult; stdcall;
function Phoneme(eCharacterSet: VOICECHARSET; dwFlags: DWORD; dText: SDATA;var dPhoneme: SDATA): HResult; stdcall;
function PosnGet(var qwTimeStamp: QWORD): HResult; stdcall;
function TextData(eCharacterSet: VOICECHARSET; dwFlags: DWORD; dText: SDATA;pNotifyInterface: Pointer; IIDNotifyInterface: TGUID): HResult; stdcall;
function ToFileTime(var qTimeStamp: QWORD; var FT: TFileTime): HResult; stdcall;
function AudioPause: HResult; stdcall;
function AudioResume: HResult; stdcall;
function AudioReset: HResult; stdcall;
function Register(pNotifyInterface: Pointer; IIDNotifyInterface: TGUID;var dwKey: DWORD): HResult; stdcall;
function UnRegister(dwKey: DWORD): HResult; stdcall;
end;
{Enumeration interface}
ITTSEnum = interface(IUnknown)
['{05EB6C6D-DBAB-11CD-B3CA-00AA0047BA4F}']
function Next(celt: ULONG; var rgelt; pceltFetched: PULONG): HResult; stdcall;
function Skip(celt: ULONG): HResult; stdcall;
function Reset: HResult; stdcall;
function Clone(out enum: ITTSEnum): HResult; stdcall;
function Select(gModeID: TGUID; Var Central: ITTSCentral;IUnknownForAudio:iUnknown): HResult; stdcall;
end;
(* А это я выдернул из SPEECH.H SAPI SDK 4.0

DEFINE_GUID(IID_IAudioMultiMediaDevice,
0xb68ad320, 0xc743, 0x11cd, 0x80, 0xe5, 0x0, 0xaa, 0x0, 0x3e, 0x4b, 0x50);

DECLARE_INTERFACE_ (IAudioMultiMediaDevice, IUnknown) {
// IUnknown members
STDMETHOD (QueryInterface) (THIS_ REFIID riid, LPVOID FAR* ppvObj) PURE;
STDMETHOD_(ULONG,AddRef) (THIS) PURE;
STDMETHOD_(ULONG,Release) (THIS) PURE;

// IAudioMultiMediaDevice members
STDMETHOD (CustomMessage) (THIS_ UINT, SDATA) PURE;
STDMETHOD (DeviceNumGet) (THIS_ DWORD* ) PURE;
STDMETHOD (DeviceNumSet) (THIS_ DWORD) PURE;
};

DEFINE_GUID(IID_ITTSEnumA, 0x05EB6C6DL, 0xDBAB, 0x11CD, 0xB3, 0xCA, 0x00, 0xAA, 0x00, 0x47, 0xBA, 0x4F);

DECLARE_INTERFACE_ (ITTSEnumA, IUnknown) {

// IUnknown members
STDMETHOD (QueryInterface) (THIS_ REFIID riid, LPVOID FAR* ppvObj) PURE;
STDMETHOD_(ULONG,AddRef) (THIS) PURE;
STDMETHOD_(ULONG,Release) (THIS) PURE;

// ITTSEnumA members
STDMETHOD (Next) (THIS_ ULONG, PTTSMODEINFOA, ULONG * ) PURE;
STDMETHOD (Skip) (THIS_ ULONG) PURE;
STDMETHOD (Reset) (THIS) PURE;
STDMETHOD (Clone) (THIS_ ITTSEnumA * FAR * ) PURE;
STDMETHOD (Select) (THIS_ GUID, PITTSCENTRALA *, LPUNKNOWN) PURE;
};
DEFINE_GUID(IID_ITTSCentralA, 0x05EB6C6AL, 0xDBAB, 0x11CD, 0xB3, 0xCA, 0x00, 0xAA, 0x00, 0x47, 0xBA, 0x4F);

DECLARE_INTERFACE_ (ITTSCentralA, IUnknown) {

// IUnknown members
STDMETHOD (QueryInterface) (THIS_ REFIID riid, LPVOID FAR* ppvObj) PURE;
STDMETHOD_(ULONG,AddRef) (THIS) PURE;
STDMETHOD_(ULONG,Release) (THIS) PURE;

// ITTSCentralA members
STDMETHOD (Inject) (THIS_ PCSTR) PURE;
STDMETHOD (ModeGet) (THIS_ PTTSMODEINFOA) PURE;
STDMETHOD (Phoneme) (THIS_ VOICECHARSET, DWORD, SDATA, PSDATA) PURE;
STDMETHOD (PosnGet) (THIS_ PQWORD) PURE;
STDMETHOD (TextData) (THIS_ VOICECHARSET, DWORD, SDATA, PVOID, IID) PURE;
STDMETHOD (ToFileTime) (THIS_ PQWORD, FILETIME * ) PURE;
STDMETHOD (AudioPause) (THIS) PURE;
STDMETHOD (AudioResume) (THIS) PURE;
STDMETHOD (AudioReset) (THIS) PURE;
STDMETHOD (Register) (THIS_ PVOID, IID, DWORD* ) PURE;
STDMETHOD (UnRegister) (THIS_ DWORD) PURE;
};

*)

VAR
fIAMM:IAudioMultimediaDevice;
aTTSEnum:ITTSEnum;
ModeInfo:TTSModeInfo;
Central:ITTSCentral;
SDat:SData;
Info:string;
I,EngineNumber:dword;
EngineCount:longint;
Error:hResult;
ID:TGUID;
N:dword;

FUNCTION SysErrorName:string;
Var
Msg:string;
L:byte absolute Msg;
begin
fillchar(Msg,sizeof(Msg),0);
L:=0;
FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM,NIL,GetLastError,0,@Msg[1],254,NIL);
while (Msg[L+1]<>#0) and (L<254) do inc(L);
if Msg[L]=#13 then dec(L);
SysErrorName:=Msg
end;

PROCEDURE Abort(Msg:string);
begin
CoUninitialize;
Msg[byte(Msg[0])]:=#0;
MessageBox(0,@Msg[1],'ERROR!',MB_OK);
halt
end;

BEGIN {MAIN}
if CoInitialize(nil) and Severity<>0 then Abort(SysErrorName); {Инициализируем COM}
if CoCreateInstance(CLSID_MMAudioDest,NIL,CLSCTX_ALL,IID_IAudioMultiMediaDevice, fIAMM)<>S_OK
then Abort('MMAudio'#13+SysErrorName); {Получаем аудио интерфейс}
if CoCreateInstance(CLSID_TTSEnumerator,NIL,CLSCTX_ALL,IID_ITTSEnum, aTTSEnum)<>S_OK
then Abort('Enumerator'#13+SysErrorName); {Получаем интерфейс нумератора}
aTTSEnum.Reset;
Info:='';
repeat {Получаем все голосовые движки}
aTTSEnum.Next(1, ModeInfo, @EngineCount);
if EngineCount<>0 then Info:=Info+StrPas(ModeInfo.szModeName)+#13
until EngineCount=0;
fIAMM.DeviceNumSet(WAVE_MAPPER); {Устанавливаем режим аудио}
aTTSEnum.Reset;
aTTSEnum.Next(1, ModeInfo, @EngineCount); {Берем 1-й движок}
Error:=aTTSEnum.Select(ModeInfo.gModeID,Central,IUnknown(fIAMM)); {Получаем центральный интерфейс}
{И АМБЕЦ - ошибка SAPI 'INVALUD MODE' или даже GENERAL PROTECTION FAULT!}

aTTSEnum.Release; {Освободить интерфейсы}
fIAMM.Release;
CoFreeUnusedLibraries; {Освободить COM}
CoUninitialize;
Abort(Info+#13+lHex(Central,4)+#13'['+lHex(Error,4)+']')
END.

Форум на Исходниках.RU

text-to-speech на Паскале

text-to-speech на Паскале