include/sapi.idl

   1 /*
   2  * Speech API (SAPI) IDL file.
   3  *
   4  * Copyright (C) 2017 Huw Davies
   5  *
   6  * This library is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * This library is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with this library; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
  19  */
  20
  21 import "oaidl.idl";
  22 import "ocidl.idl";
  23
  24 #ifndef __WIDL__
  25 #define threading(model)
  26 #define progid(str)
  27 #define vi_progid(str)
  28 #endif
  29
  30 cpp_quote("#if 0")
  31 typedef [restricted, hidden] struct WAVEFORMATEX
  32 {
  33     WORD  wFormatTag;
  34     WORD  nChannels;
  35     DWORD nSamplesPerSec;
  36     DWORD nAvgBytesPerSec;
  37     WORD  nBlockAlign;
  38     WORD  wBitsPerSample;
  39     WORD  cbSize;
  40 } WAVEFORMATEX;
  41 cpp_quote("#else")
  42 cpp_quote("#include <mmsystem.h>")
  43 cpp_quote("#endif")
  44
  45 cpp_quote("#if defined(__cplusplus)")
  46 cpp_quote("interface ISpNotifyCallback")
  47 cpp_quote("{")
  48 cpp_quote("virtual HRESULT STDMETHODCALLTYPE NotifyCallback(WPARAM wParam, LPARAM lParam) = 0;")
  49 cpp_quote("};")
  50 cpp_quote("#else")
  51 typedef void * ISpNotifyCallback;
  52 cpp_quote("#endif")
  53
  54 cpp_quote("#if 0")
  55 typedef void * SPNOTIFYCALLBACK;
  56 typedef void * SPSTATEHANDLE;
  57 cpp_quote("#endif")
  58
  59 typedef [restricted, hidden] struct SPEVENT
  60 {
  61     WORD      eEventId;
  62     WORD      elParamType;
  63     ULONG     ulStreamNum;
  64     ULONGLONG ullAudioStreamOffset;
  65     WPARAM    wParam;
  66     LPARAM    lParam;
  67 } SPEVENT;
  68
  69 cpp_quote("typedef void __stdcall SPNOTIFYCALLBACK(WPARAM wParam, LPARAM lParam);")
  70 cpp_quote("DECLARE_HANDLE(SPSTATEHANDLE);")
  71
  72 const ULONG SP_MAX_LANGIDS = 20;
  73
  74 typedef [restricted, hidden] struct SPEVENTSOURCEINFO
  75 {
  76     ULONGLONG   ullEventInterest;
  77     ULONGLONG   ullQueuedInterest;
  78     ULONG       ulCount;
  79 } SPEVENTSOURCEINFO;
  80
  81 typedef [hidden] enum SPGRAMMARWORDTYPE
  82 {
  83     SPWT_DISPLAY,
  84     SPWT_LEXICAL,
  85     SPWT_PRONUNCIATION,
  86     SPWT_LEXICAL_NO_SPECIAL_CHARS
  87 } SPGRAMMARWORDTYPE;
  88
  89 typedef [hidden] struct tagSPPROPERTYINFO
  90 {
  91     const WCHAR *pszName;
  92     ULONG        ulId;
  93     const WCHAR *pszValue;
  94     VARIANT      vValue;
  95 } SPPROPERTYINFO;
  96
  97 typedef [hidden] enum SPDATAKEYLOCATION
  98 {
  99     SPDKL_DefaultLocation = 0,
 100     SPDKL_CurrentUser     = 1,
 101     SPDKL_LocalMachine    = 2,
 102     SPDKL_CurrentConfig   = 5
 103 } SPDATAKEYLOCATION;
 104
 105 typedef [hidden] enum SPLOADOPTIONS
 106 {
 107     SPLO_STATIC,
 108     SPLO_DYNAMIC
 109 } SPLOADOPTIONS;
 110
 111 typedef [hidden] enum SPRULESTATE
 112 {
 113     SPRS_INACTIVE               = 0,
 114     SPRS_ACTIVE                 = 1,
 115     SPRS_ACTIVE_WITH_AUTO_PAUSE = 3,
 116     SPRS_ACTIVE_USER_DELIMITED  = 4
 117 } SPRULESTATE;
 118
 119 typedef [hidden] enum SPGRAMMARSTATE
 120 {
 121     SPGS_DISABLED  = 0,
 122     SPGS_ENABLED   = 1,
 123     SPGS_EXCLUSIVE = 3
 124 } SPGRAMMARSTATE;
 125
 126 typedef [hidden] struct SPBINARYGRAMMAR
 127 {
 128     ULONG ulTotalSerializedSize;
 129 } SPBINARYGRAMMAR;
 130
 131 typedef [hidden] struct tagSPTEXTSELECTIONINFO
 132 {
 133     ULONG ulStartActiveOffset;
 134     ULONG cchActiveChars;
 135     ULONG ulStartSelection;
 136     ULONG cchSelection;
 137 } SPTEXTSELECTIONINFO;
 138
 139 typedef [hidden] enum SPWORDPRONOUNCEABLE
 140 {
 141     SPWP_UNKNOWN_WORD_UNPRONOUNCEABLE,
 142     SPWP_UNKNOWN_WORD_PRONOUNCEABLE,
 143     SPWP_KNOWN_WORD_PRONOUNCEABLE
 144 } SPWORDPRONOUNCEABLE;
 145
 146 typedef [hidden] enum SPINTERFERENCE
 147 {
 148     SPINTERFERENCE_NONE,
 149     SPINTERFERENCE_NOISE,
 150     SPINTERFERENCE_NOSIGNAL,
 151     SPINTERFERENCE_TOOLOUD,
 152     SPINTERFERENCE_TOOQUIET,
 153     SPINTERFERENCE_TOOFAST,
 154     SPINTERFERENCE_TOOSLOW
 155 } SPINTERFERENCE;
 156
 157 typedef [hidden] enum SPVISEMES
 158 {
 159     SP_VISEME_0 = 0,
 160     SP_VISEME_1,
 161     SP_VISEME_2,
 162     SP_VISEME_3,
 163     SP_VISEME_4,
 164     SP_VISEME_5,
 165     SP_VISEME_6,
 166     SP_VISEME_7,
 167     SP_VISEME_8,
 168     SP_VISEME_9,
 169     SP_VISEME_10,
 170     SP_VISEME_11,
 171     SP_VISEME_12,
 172     SP_VISEME_13,
 173     SP_VISEME_14,
 174     SP_VISEME_15,
 175     SP_VISEME_16,
 176     SP_VISEME_17,
 177     SP_VISEME_18,
 178     SP_VISEME_19,
 179     SP_VISEME_20,
 180     SP_VISEME_21,
 181 } SPVISEMES;
 182
 183 typedef [hidden] enum SPEVENTENUM
 184 {
 185     SPEI_UNDEFINED           = 0,
 186     SPEI_START_INPUT_STREAM  = 1,
 187     SPEI_END_INPUT_STREAM    = 2,
 188     SPEI_VOICE_CHANGE        = 3,
 189     SPEI_TTS_BOOKMARK        = 4,
 190     SPEI_WORD_BOUNDARY       = 5,
 191     SPEI_PHONEME             = 6,
 192     SPEI_SENTENCE_BOUNDARY   = 7,
 193     SPEI_VISEME              = 8,
 194     SPEI_TTS_AUDIO_LEVEL     = 9,
 195     SPEI_TTS_PRIVATE         = 15,
 196
 197     SPEI_MIN_TTS             = 1,
 198     SPEI_MAX_TTS             = 15,
 199
 200     SPEI_END_SR_STREAM          = 34,
 201     SPEI_SOUND_START            = 35,
 202     SPEI_SOUND_END              = 36,
 203     SPEI_PHRASE_START           = 37,
 204     SPEI_RECOGNITION            = 38,
 205     SPEI_HYPOTHESIS             = 39,
 206     SPEI_SR_BOOKMARK            = 40,
 207     SPEI_PROPERTY_NUM_CHANGE    = 41,
 208     SPEI_PROPERTY_STRING_CHANGE = 42,
 209     SPEI_FALSE_RECOGNITION      = 43,
 210     SPEI_INTERFERENCE           = 44,
 211     SPEI_REQUEST_UI             = 45,
 212     SPEI_RECO_STATE_CHANGE      = 46,
 213     SPEI_ADAPTATION             = 47,
 214     SPEI_START_SR_STREAM        = 48,
 215     SPEI_RECO_OTHER_CONTEXT     = 49,
 216     SPEI_SR_AUDIO_LEVEL         = 50,
 217     SPEI_SR_RETAINEDAUDIO       = 51,
 218     SPEI_SR_PRIVATE             = 52,
 219     SPEI_ACTIVE_CATEGORY_CHANGED = 53,
 220
 221     SPEI_MIN_SR              = 34,
 222     SPEI_MAX_SR              = 55,
 223
 224     SPEI_RESERVED1           = 30,
 225     SPEI_RESERVED2           = 33,
 226     SPEI_RESERVED3           = 63
 227 } SPEVENTENUM;
 228
 229 typedef [hidden] enum SPCONTEXTSTATE
 230 {
 231     SPCS_DISABLED = 0,
 232     SPCS_ENABLED  = 1,
 233 } SPCONTEXTSTATE;
 234
 235 typedef [hidden] enum SPVPRIORITY
 236 {
 237     SPVPRI_NORMAL = 0,
 238     SPVPRI_ALERT  = 1,
 239     SPVPRI_OVER   = 2
 240 } SPVPRIORITY;
 241
 242 typedef [hidden] enum SPRECOSTATE
 243 {
 244     SPRST_INACTIVE,
 245     SPRST_ACTIVE,
 246     SPRST_ACTIVE_ALWAYS,
 247     SPRST_INACTIVE_WITH_PURGE,
 248     SPRST_NUM_STATES
 249 } SPRECOSTATE;
 250
 251 typedef [hidden] enum SPWAVEFORMATTYPE
 252 {
 253     SPWF_INPUT,
 254     SPWF_SRENGINE
 255 } SPSTREAMFORMATTYPE;
 256
 257 typedef [restricted, hidden] struct SPRECOCONTEXTSTATUS
 258 {
 259     SPINTERFERENCE  eInterference;
 260     WCHAR           szRequestTypeOfUI[255];
 261     DWORD           dwReserved1;
 262     DWORD           dwReserved2;
 263 } SPRECOCONTEXTSTATUS;
 264
 265 typedef [hidden] enum SPAUDIOOPTIONS
 266 {
 267     SPAO_NONE,
 268     SPAO_RETAIN_AUDIO
 269 } SPAUDIOOPTIONS;
 270
 271 typedef [hidden] struct SPSERIALIZEDRESULT
 272 {
 273     ULONG ulSerializedSize;
 274 } SPSERIALIZEDRESULT;
 275
 276 typedef [hidden] enum SPDISPLYATTRIBUTES
 277 {
 278     SPAF_ONE_TRAILING_SPACE     = 0x02,
 279     SPAF_TWO_TRAILING_SPACES    = 0x04,
 280     SPAF_CONSUME_LEADING_SPACES = 0x08,
 281     SPAF_ALL                    = 0x0f
 282 } SPDISPLAYATTRIBUTES;
 283
 284 typedef [hidden] enum SPBOOKMARKOPTIONS
 285 {
 286     SPBO_NONE,
 287     SPBO_PAUSE
 288 } SPBOOKMARKOPTIONS;
 289
 290 typedef [hidden] enum _SPAUDIOSTATE
 291 {
 292     SPAS_CLOSED,
 293     SPAS_STOP,
 294     SPAS_PAUSE,
 295     SPAS_RUN
 296 } SPAUDIOSTATE;
 297
 298 typedef [hidden] enum SPRUNSTATE
 299 {
 300     SPRS_DONE        = (1L << 0),
 301     SPRS_IS_SPEAKING = (1L << 1)
 302 } SPRUNSTATE;
 303
 304 typedef [hidden] enum SPEAKFLAGS
 305 {
 306     SPF_DEFAULT            = 0,
 307     SPF_ASYNC              = (1L << 0),
 308     SPF_PURGEBEFORESPEAK   = (1L << 1),
 309     SPF_IS_FILENAME        = (1L << 2),
 310     SPF_IS_XML             = (1L << 3),
 311     SPF_IS_NOT_XML         = (1L << 4),
 312     SPF_PERSIST_XML        = (1L << 5),
 313     SPF_NLP_SPEAK_PUNC     = (1L << 6),
 314     SPF_PARSE_SAPI         = (1L << 7),
 315     SPF_PARSE_SSML         = (1L << 8),
 316     SPF_PARSE_AUTODETECT   = 0,
 317     SPF_NLP_MASK           = (SPF_NLP_SPEAK_PUNC),
 318     SPF_PARSE_MASK         = (SPF_PARSE_SAPI|SPF_PARSE_SSML),
 319     SPF_VOICE_MASK         = (SPF_ASYNC|SPF_PURGEBEFORESPEAK|SPF_IS_FILENAME|SPF_IS_XML|SPF_IS_NOT_XML|SPF_NLP_MASK|SPF_PERSIST_XML|SPF_PARSE_MASK),
 320     SPF_UNUSED_FLAGS       = ~(SPF_VOICE_MASK)
 321 } SPEAKFLAGS;
 322
 323 typedef [hidden] enum SPCFGRULEATTRIBUTES
 324 {
 325     SPRAF_TopLevel      = (1 << 0),
 326     SPRAF_Active        = (1 << 1),
 327     SPRAF_Export        = (1 << 2),
 328     SPRAF_Import        = (1 << 3),
 329     SPRAF_Interpreter   = (1 << 4),
 330     SPRAF_Dynamic       = (1 << 5),
 331     SPRAF_Root          = (1 << 6),
 332     SPRAF_AutoPause     = (1 << 16),
 333     SPRAF_UserDelimited = (1 << 17)
 334 } SPCFGRULEATTRIBUTES;
 335
 336 typedef [hidden] enum SPPHRASERNG
 337 {
 338     SPPR_ALL_ELEMENTS = -1
 339 } SPPHRASERNG;
 340
 341 typedef [hidden] enum SPVALUETYPE
 342 {
 343     SPDF_PROPERTY      = 0x00000001,
 344     SPDF_REPLACEMENT   = 0x00000002,
 345     SPDF_RULE          = 0x00000004,
 346     SPDF_DISPLAYTEXT   = 0x00000008,
 347     SPDF_LEXICALFORM   = 0x00000010,
 348     SPDF_PRONUNCIATION = 0x00000020,
 349     SPDF_AUDIO         = 0x00000040,
 350     SPDF_ALTERNATES    = 0x00000080,
 351     SPDF_ALL           = 0x000000ff
 352 } SPVALUETYPE;
 353
 354 typedef unsigned short SPPHONEID;
 355
 356 typedef [restricted, hidden] struct SPPHRASEELEMENT
 357 {
 358     ULONG           ulAudioTimeOffset;
 359     ULONG           ulAudioSizeTime;
 360     ULONG           ulAudioStreamOffset;
 361     ULONG           ulAudioSizeBytes;
 362     ULONG           ulRetainedStreamOffset;
 363     ULONG           ulRetainedSizeBytes;
 364     const WCHAR     *pszDisplayText;
 365     const WCHAR     *pszLexicalForm;
 366     const SPPHONEID *pszPronunciation;
 367     BYTE            bDisplayAttributes;
 368     signed char     RequiredConfidence;
 369     signed char     ActualConfidence;
 370     BYTE            Reserved;
 371     float           SREngineConfidence;
 372 } SPPHRASEELEMENT;
 373
 374 typedef [restricted, hidden] struct SPVOICESTATUS
 375 {
 376     ULONG     ulCurrentStream;
 377     ULONG     ulLastStreamQueued;
 378     HRESULT   hrLastResult;
 379     DWORD     dwRunningState;
 380     ULONG     ulInputWordPos;
 381     ULONG     ulInputWordLen;
 382     ULONG     ulInputSentPos;
 383     ULONG     ulInputSentLen;
 384     LONG      lBookmarkId;
 385     SPPHONEID PhonemeId;
 386     SPVISEMES VisemeId;
 387     DWORD     dwReserved1;
 388     DWORD     dwReserved2;
 389 } SPVOICESTATUS;
 390
 391 typedef [restricted, hidden] struct SPAUDIOSTATUS
 392 {
 393     long         cbFreeBuffSpace;
 394     ULONG        cbNonBlockingIO;
 395     SPAUDIOSTATE State;
 396     ULONGLONG    CurSeekPos;
 397     ULONGLONG    CurDevicePos;
 398     DWORD        dwAudioLevel;
 399     DWORD        dwReserved2;
 400 } SPAUDIOSTATUS;
 401
 402 typedef [restricted, hidden] struct SPRECOGNIZERSTATUS
 403 {
 404     SPAUDIOSTATUS AudioStatus;
 405     ULONGLONG     ullRecognitionStreamPos;
 406     ULONG         ulStreamNumber;
 407     ULONG         ulNumActive;
 408     CLSID         clsidEngine;
 409     ULONG         cLangIDs;
 410     LANGID        aLangID[SP_MAX_LANGIDS];
 411     ULONGLONG     ullRecognitionStreamTime;
 412 } SPRECOGNIZERSTATUS;
 413
 414 typedef [restricted, hidden] struct SPPHRASERULE SPPHRASERULE;
 415
 416 struct SPPHRASERULE
 417 {
 418     LPCWSTR                 pszName;
 419     ULONG                   ulId;
 420     ULONG                   ulFirstElement;
 421     ULONG                   ulCountOfElements;
 422     const SPPHRASERULE*     pNextSibling;
 423     const SPPHRASERULE*     pFirstChild;
 424     float                   SREngineConfidence;
 425     signed char             Confidence;
 426 };
 427
 428 typedef [restricted, hidden] struct SPPHRASEPROPERTY SPPHRASEPROPERTY;
 429
 430 struct SPPHRASEPROPERTY
 431 {
 432     const WCHAR            *pszName;
 433     ULONG                   ulId;
 434     const WCHAR            *pszValue;
 435     VARIANT                 vValue;
 436     ULONG                   ulFirstElement;
 437     ULONG                   ulCountOfElements;
 438     const SPPHRASEPROPERTY *pNextSibling;
 439     const SPPHRASEPROPERTY *pFirstChild;
 440     float                   SREngineConfidence;
 441     signed char             Confidence;
 442 };
 443
 444 typedef [restricted, hidden] struct SPPHRASEREPLACEMENT
 445 {
 446     BYTE         bDisplayAttributes;
 447     const WCHAR *pszReplacementText;
 448     ULONG        ulFirstElement;
 449     ULONG        ulCountOfElements;
 450 } SPPHRASEREPLACEMENT;
 451
 452 typedef [restricted, hidden] struct SPPHRASE
 453 {
 454     ULONG                   cbSize;
 455     LANGID                  LangID;
 456     WORD                    wReserved;
 457     ULONGLONG               ullGrammarID;
 458     ULONGLONG               ftStartTime;
 459     ULONGLONG               ullAudioStreamPosition;
 460     ULONG                   ulAudioSizeBytes;
 461     ULONG                   ulRetainedSizeBytes;
 462     ULONG                   ulAudioSizeTime;
 463     SPPHRASERULE            Rule;
 464     const SPPHRASEPROPERTY *pProperties;
 465     const SPPHRASEELEMENT  *pElements;
 466     ULONG                   cReplacements;
 467     const SPPHRASEREPLACEMENT *pReplacements;
 468     GUID                    SREngineID;
 469     ULONG                   ulSREnginePrivateDataSize;
 470     const BYTE             *pSREnginePrivateData;
 471 } SPPHRASE;
 472
 473 typedef [restricted, hidden] struct SPSERIALIZEDPHRASE
 474 {
 475     ULONG ulSerializedSize;
 476 } SPSERIALIZEDPHRASE;
 477
 478 typedef [restricted, hidden] struct SPRECORESULTTIMES
 479 {
 480     FILETIME  ftStreamTime;
 481     ULONGLONG ullLength;
 482     DWORD     dwTickCount;
 483     ULONGLONG ullStart;
 484 } SPRECORESULTTIMES;
 485
 486 typedef [hidden] enum SPFILEMODE
 487 {
 488     SPFM_OPEN_READONLY,
 489     SPFM_OPEN_READWRITE,
 490     SPFM_CREATE,
 491     SPFM_CREATE_ALWAYS,
 492     SPFM_NUM_MODES
 493 } SPFILEMODE;
 494
 495 cpp_quote("#if defined(__GNUC__)")
 496
 497 cpp_quote("#define SPCAT_AUDIOOUT (const WCHAR []){ 'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','A','u','d','i','o','O','u','t','p','u','t',0 }")
 498 cpp_quote("#define SPCAT_AUDIOIN (const WCHAR []){ 'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','A','u','d','i','o','I','n','p','u','t',0 }")
 499 cpp_quote("#define SPCAT_VOICES (const WCHAR []){ 'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','V','o','i','c','e','s',0 }")
 500
 501 cpp_quote("#elif defined(_MSC_VER)")
 502
 503 cpp_quote("#define SPCAT_AUDIOOUT L\"HKEY_LOCAL_MACHINE\\\\SOFTWARE\\\\Microsoft\\\\Speech\\\\AudioOutput\"")
 504 cpp_quote("#define SPCAT_AUDIOIN L\"HKEY_LOCAL_MACHINE\\\\SOFTWARE\\\\Microsoft\\\\Speech\\\\AudioInput\"")
 505 cpp_quote("#define SPCAT_VOICES L\"HKEY_LOCAL_MACHINE\\\\SOFTWARE\\\\Microsoft\\\\Speech\\\\Voices\"")
 506
 507 cpp_quote("#else")
 508
 509 cpp_quote("static const WCHAR SPCAT_AUDIOOUT[] = {'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','A','u','d','i','o','O','u','t','p','u','t',0};")
 510 cpp_quote("static const WCHAR SPCAT_AUDIOIN[] = {'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','A','u','d','i','o','I','n','p','u','t',0};")
 511 cpp_quote("static const WCHAR SPCAT_VOICES[] = {'H','K','E','Y','_','L','O','C','A','L','_','M','A','C','H','I','N','E','\\\\','S','O','F','T','W','A','R','E','\\\\','M','i','c','r','o','s','o','f','t','\\\\','S','p','e','e','c','h','\\\\','V','o','i','c','e','s',0};")
 512
 513 cpp_quote("#endif")
 514
 515 interface IEnumSpObjectTokens;
 516 interface ISpEventSource;
 517 interface ISpRecoContext;
 518 interface ISpRecognizer;
 519 interface ISpStreamFormat;
 520
 521 [
 522     object,
 523     uuid(14056581-e16c-11d2-bb90-00c04f8ee6c0),
 524     helpstring("ISpDataKey"),
 525     pointer_default(unique),
 526     local,
 527     restricted
 528 ]
 529 interface ISpDataKey : IUnknown
 530 {
 531     HRESULT SetData([in] LPCWSTR pszValueName,
 532                     [in] ULONG cbData,
 533                     [in] const BYTE *pData);
 534     HRESULT GetData([in] LPCWSTR pszValueName,
 535                     [in] ULONG *pcbData,
 536                     [out] BYTE *pData);
 537     HRESULT SetStringValue([in] LPCWSTR pszValueName,
 538                            [in] LPCWSTR pszValue);
 539     HRESULT GetStringValue([in] LPCWSTR pszValueName,
 540                            [out] LPWSTR *ppszValue);
 541     HRESULT SetDWORD([in] LPCWSTR pszValueName,
 542                      [in] DWORD dwValue);
 543     HRESULT GetDWORD([in] LPCWSTR pszValueName,
 544                      [out] DWORD *pdwValue);
 545     HRESULT OpenKey([in] LPCWSTR pszSubKeyName,
 546                     [out] ISpDataKey **ppSubKey);
 547     HRESULT CreateKey([in] LPCWSTR pszSubKey,
 548                       [out] ISpDataKey **ppSubKey);
 549     HRESULT DeleteKey([in] LPCWSTR pszSubKey);
 550     HRESULT DeleteValue([in] LPCWSTR pszValueName);
 551     HRESULT EnumKeys([in] ULONG Index,
 552                      [out] LPWSTR *ppszSubKeyName);
 553     HRESULT EnumValues([in] ULONG Index,
 554                        [out] LPWSTR *ppszValueName);
 555 }
 556 [
 557     object,
 558     uuid(92a66e2b-c830-4149-83df-6fc2ba1e7a5b),
 559     helpstring("ISpRegDataKey"),
 560     pointer_default(unique),
 561     restricted
 562 ]
 563 interface ISpRegDataKey : ISpDataKey
 564 {
 565     [local] HRESULT SetKey([in] HKEY hkey, [in] BOOL fReadOnly);
 566 }
 567
 568 [
 569     object,
 570     uuid(2d3d3845-39af-4850-bbf9-40b49780011d),
 571     helpstring("ISpObjectTokenCategory"),
 572     pointer_default(unique),
 573     local,
 574     restricted
 575 ]
 576 interface ISpObjectTokenCategory : ISpDataKey
 577 {
 578     HRESULT SetId([in] LPCWSTR pszCategoryId,
 579                   [in] BOOL fCreateIfNotExist);
 580     HRESULT GetId([out] LPWSTR *ppszCoMemCategoryId);
 581     HRESULT GetDataKey([in] SPDATAKEYLOCATION spdkl,
 582                        [out] ISpDataKey **ppDataKey);
 583     HRESULT EnumTokens([in, string] LPCWSTR pszReqAttribs,
 584                        [in, string] LPCWSTR pszOptAttribs,
 585                        [out] IEnumSpObjectTokens **ppEnum);
 586     HRESULT SetDefaultTokenId([in] LPCWSTR pszTokenId);
 587     HRESULT GetDefaultTokenId([out] LPWSTR *ppszCoMemTokenId);
 588 }
 589
 590 [
 591     object,
 592     uuid(14056589-e16c-11d2-bb90-00c04f8ee6c0),
 593     helpstring("ISpObjectToken"),
 594     pointer_default(unique),
 595     local,
 596     restricted
 597 ]
 598 interface ISpObjectToken : ISpDataKey
 599 {
 600     HRESULT SetId(/*[in]*/ LPCWSTR pszCategoryId,
 601                   [in] LPCWSTR pszTokenId,
 602                   [in] BOOL fCreateIfNotExist);
 603     HRESULT GetId([out] LPWSTR *ppszCoMemTokenId);
 604     HRESULT GetCategory([out] ISpObjectTokenCategory **ppTokenCategory);
 605     HRESULT CreateInstance([in] IUnknown *pUnkOuter,
 606                            [in] DWORD dwClsContext,
 607                            [in] REFIID riid,
 608                            [out, iid_is(riid)] void **ppvObject);
 609     HRESULT GetStorageFileName([in] REFCLSID clsidCaller,
 610                                [in] LPCWSTR pszValueName,
 611                                [in, string] LPCWSTR pszFileNameSpecifier,
 612                                [in] ULONG nFolder,
 613                                [out] LPWSTR *ppszFilePath);
 614     HRESULT RemoveStorageFileName([in] REFCLSID clsidCaller,
 615                                   [in] LPCWSTR pszKeyName,
 616                                   [in] BOOL fDeleteFile);
 617     HRESULT Remove(/*[in]*/ const CLSID *pclsidCaller);
 618     [local] HRESULT IsUISupported([in] LPCWSTR pszTypeOfUI,
 619                                   [in] void *pvExtraData,
 620                                   [in] ULONG cbExtraData,
 621                                   [in] IUnknown *punkObject,
 622                                   [out] BOOL *pfSupported);
 623     [local] HRESULT DisplayUI([in] HWND hwndParent,
 624                               [in] LPCWSTR pszTitle,
 625                               [in] LPCWSTR pszTypeOfUI,
 626                               [in] void *pvExtraData,
 627                               [in] ULONG cbExtraData,
 628                               [in] IUnknown *punkObject);
 629     HRESULT MatchesAttributes([in] LPCWSTR pszAttributes,
 630                               [out] BOOL *pfMatches);
 631 }
 632
 633 [
 634     object,
 635     uuid(06b64f9e-7fda-11d2-b4f2-00c04f797396),
 636     helpstring("IEnumSpObjectTokens"),
 637     pointer_default(unique),
 638     local,
 639     restricted
 640 ]
 641 interface IEnumSpObjectTokens : IUnknown
 642 {
 643     HRESULT Next([in] ULONG celt,
 644                  [out, size_is(celt), length_is(*pceltFetched)] ISpObjectToken **pelt,
 645                  [out] ULONG *pceltFetched);
 646     HRESULT Skip([in] ULONG celt);
 647     HRESULT Reset(void);
 648     HRESULT Clone([out] IEnumSpObjectTokens **ppEnum);
 649     HRESULT Item([in] ULONG Index,
 650                  [out] ISpObjectToken **ppToken);
 651     HRESULT GetCount([out] ULONG *pCount);
 652 }
 653
 654 [
 655     object,
 656     uuid(1a5c0354-b621-4b5a-8791-d306ed379e53),
 657     local,
 658     pointer_default(unique),
 659     restricted
 660 ]
 661 interface ISpPhrase : IUnknown
 662 {
 663     HRESULT GetPhrase([out] SPPHRASE **phrase);
 664     HRESULT GetSerializedPhrase([out] SPSERIALIZEDPHRASE **phrase);
 665     HRESULT GetText([in] ULONG start, [in] ULONG count, [in] BOOL replacements,
 666                                         [out] WCHAR **text, [out] BYTE *attributes);
 667     HRESULT Discard([in] DWORD types);
 668 };
 669
 670 [
 671     object,
 672     uuid(8fcebc98-4e49-4067-9c6c-d86a0e092e3d),
 673     local,
 674     pointer_default(unique),
 675     restricted
 676 ]
 677 interface ISpPhraseAlt : ISpPhrase
 678 {
 679     HRESULT GetAltInfo(ISpPhrase **parent, ULONG *start, ULONG *count,
 680                         ULONG *elements);
 681     HRESULT Commit();
 682 };
 683
 684 [
 685     object,
 686     uuid(20b053be-e235-43cd-9a2a-8d17a48b7842),
 687     pointer_default(unique),
 688     local,
 689     restricted
 690 ]
 691 interface ISpRecoResult : ISpPhrase
 692 {
 693     HRESULT GetResultTimes([out] SPRECORESULTTIMES *times);
 694
 695     HRESULT GetAlternates(
 696                 [in] ULONG start,
 697                 [in] ULONG elements,
 698                 [in] ULONG count,
 699                 [out] ISpPhraseAlt **phrases,
 700                 [out] ULONG *returned);
 701
 702     HRESULT GetAudio(
 703                 [in] ULONG start,
 704                 [in] ULONG elements,
 705                 [out] ISpStreamFormat **stream);
 706     HRESULT SpeakAudio(
 707                 [in] ULONG start,
 708                 [in] ULONG elements,
 709                 [in] DWORD flags,
 710                 [out] ULONG *number);
 711
 712     HRESULT Serialize([out] SPSERIALIZEDRESULT **result);
 713
 714     HRESULT ScaleAudio([in] const GUID *format, [in] const WAVEFORMATEX *wave);
 715
 716     HRESULT GetRecoContext([out] ISpRecoContext *context);
 717 };
 718
 719 [
 720     object,
 721     uuid(8137828f-591a-4a42-be58-49ea7ebaac68),
 722     helpstring("ISpGrammarBuilder Interface"),
 723     pointer_default(unique),
 724     restricted,
 725     local
 726 ]
 727 interface ISpGrammarBuilder : IUnknown
 728 {
 729     HRESULT ResetGrammar([in] LANGID language);
 730
 731     HRESULT GetRule([in] const WCHAR *rule,
 732                     [in] DWORD id,
 733                     [in] DWORD attributes,
 734                     [in] BOOL should_create,
 735                     [out] SPSTATEHANDLE *state);
 736     HRESULT ClearRule(SPSTATEHANDLE state);
 737
 738     HRESULT CreateNewState(
 739                 [in] SPSTATEHANDLE instate,
 740                 [out] SPSTATEHANDLE *outstate);
 741
 742     HRESULT AddWordTransition(
 743                     [in] SPSTATEHANDLE from_state,
 744                     [in] SPSTATEHANDLE to_state,
 745                     [in] const WCHAR *value,
 746                     [in] const WCHAR *separators,
 747                     [in] SPGRAMMARWORDTYPE type,
 748                     [in] float weight,
 749                     [in] const SPPROPERTYINFO *info);
 750
 751     HRESULT AddRuleTransition(
 752                 [in] SPSTATEHANDLE from_state,
 753                 [in] SPSTATEHANDLE to_state,
 754                 [in] SPSTATEHANDLE rule,
 755                 [in] float weight,
 756                 [in] const SPPROPERTYINFO *info);
 757
 758     HRESULT AddResource([in] SPSTATEHANDLE state,
 759                         [in] const WCHAR *name,
 760                         [in] const WCHAR *value);
 761
 762     HRESULT Commit([in] DWORD reserved);
 763 };
 764
 765 [
 766     object,
 767     uuid(2177db29-7f45-47d0-8554-067e91c80502),
 768     pointer_default(unique),
 769     restricted,
 770     local
 771 ]
 772 interface ISpRecoGrammar : ISpGrammarBuilder
 773 {
 774     HRESULT GetGrammarId([out] ULONGLONG *grammar);
 775     HRESULT GetRecoContext([out] ISpRecoContext **context);
 776
 777     HRESULT LoadCmdFromFile([in, string] const WCHAR * pszFileName, [in] SPLOADOPTIONS options);
 778     HRESULT LoadCmdFromObject([in] REFCLSID rcid, [in, string] const WCHAR * pszGrammarName, [in] SPLOADOPTIONS options);
 779     HRESULT LoadCmdFromResource([in] HMODULE hmod,
 780                                 [in, string] const WCHAR *name,
 781                                 [in, string] const WCHAR *type,
 782                                 [in] WORD language,
 783                                 [in] SPLOADOPTIONS options);
 784     HRESULT LoadCmdFromMemory([in] const SPBINARYGRAMMAR *grammar, [in] SPLOADOPTIONS options);
 785     HRESULT LoadCmdFromProprietaryGrammar(
 786                                 [in] REFGUID guid,
 787                                 [in, string] const WCHAR *param,
 788                                 [in] const void *data,
 789                                 [in] ULONG size,
 790                                 [in] SPLOADOPTIONS options);
 791     HRESULT SetRuleState([in, string] const WCHAR *name, [in] void *reserved,
 792                          [in] SPRULESTATE state);
 793     HRESULT SetRuleIdState([in] ULONG rule, [in] SPRULESTATE state);
 794
 795     HRESULT LoadDictation([in, string] const WCHAR *name, [in] SPLOADOPTIONS options);
 796     HRESULT UnloadDictation();
 797     HRESULT SetDictationState([in] SPRULESTATE state);
 798
 799     HRESULT SetWordSequenceData([in] const WCHAR *text, [in] ULONG count, [in] const SPTEXTSELECTIONINFO *info);
 800     HRESULT SetTextSelection([in] const SPTEXTSELECTIONINFO *info);
 801
 802     HRESULT IsPronounceable([in, string] const WCHAR *word, [out] SPWORDPRONOUNCEABLE *pronounceable);
 803
 804     HRESULT SetGrammarState([in] SPGRAMMARSTATE state);
 805
 806     HRESULT SaveCmd([in] IStream *stream, [out, optional] WCHAR **error);
 807
 808     HRESULT GetGrammarState([out] SPGRAMMARSTATE *state);
 809 };
 810
 811 [
 812     object,
 813     uuid(259684dc-37c3-11d2-9603-00c04f8ee628),
 814     pointer_default(unique),
 815     restricted
 816 ]
 817 interface ISpNotifySink : IUnknown
 818 {
 819     HRESULT Notify(void);
 820 }
 821
 822 [
 823     object,
 824     uuid(6c44df74-72b9-4992-a1ec-ef996e0422d4),
 825     pointer_default(unique),
 826     restricted
 827 ]
 828 interface ISpVoice : ISpEventSource
 829 {
 830     HRESULT SetOutput([in] IUnknown *unk, [in] BOOL changes);
 831     HRESULT GetOutputObjectToken([out] ISpObjectToken **token);
 832     HRESULT GetOutputStream([out] ISpStreamFormat **stream);
 833
 834     HRESULT Pause(void);
 835     HRESULT Resume(void);
 836
 837     HRESULT SetVoice([in] ISpObjectToken *token);
 838     HRESULT GetVoice([out] ISpObjectToken **token);
 839
 840     HRESULT Speak([in, string] const WCHAR *pwcs,
 841                 [in] DWORD flags,
 842                 [out] ULONG *number);
 843     HRESULT SpeakStream([in] IStream *stream,
 844                 [in] DWORD flags,
 845                 [out] ULONG *number);
 846
 847     HRESULT GetStatus(
 848                 [out] SPVOICESTATUS *status,
 849                 [out, string] WCHAR **bookmark);
 850
 851     HRESULT Skip([in,string] const WCHAR *type, [in] long items, [out] ULONG *skipped);
 852
 853     HRESULT SetPriority([in] SPVPRIORITY priority);
 854     HRESULT GetPriority([out] SPVPRIORITY* priority);
 855
 856     HRESULT SetAlertBoundary([in] SPEVENTENUM boundary);
 857     HRESULT GetAlertBoundary([out] SPEVENTENUM* boundary);
 858
 859     HRESULT SetRate([in] long adjust);
 860     HRESULT GetRate([out] long *adjust);
 861
 862     HRESULT SetVolume([in] USHORT volume);
 863     HRESULT GetVolume([out] USHORT *volume);
 864
 865     HRESULT WaitUntilDone([in] ULONG timeout);
 866
 867     HRESULT SetSyncSpeakTimeout([in] ULONG timeout);
 868     HRESULT GetSyncSpeakTimeout([out] ULONG *timeout);
 869
 870     [local] HANDLE SpeakCompleteEvent();
 871
 872     [local] HRESULT IsUISupported(
 873                 [in] const WCHAR *type,
 874                 [in] void * extra,
 875                 [in] ULONG count,
 876                 [out] BOOL *supported);
 877     [local] HRESULT DisplayUI(
 878                 [in] HWND parent,
 879                 [in] const WCHAR *title,
 880                 [in] const WCHAR *type,
 881                 [in] void *extra,
 882                 [in] ULONG count);
 883 };
 884
 885 [
 886     object,
 887     uuid(5eff4aef-8487-11d2-961c-00c04f8ee628),
 888     pointer_default(unique),
 889     restricted
 890 ]
 891 interface ISpNotifySource : IUnknown
 892 {
 893     HRESULT SetNotifySink([in] ISpNotifySink *sink);
 894     [local] HRESULT SetNotifyWindowMessage(
 895                 [in] HWND hwnd,
 896                 [in] UINT msg,
 897                 [in] WPARAM wparam,
 898                 [in] LPARAM lparam);
 899     [local] HRESULT SetNotifyCallbackFunction(
 900                 [in] SPNOTIFYCALLBACK *callback,
 901                 [in] WPARAM wparam,
 902                 [in] LPARAM lparam);
 903     [local] HRESULT SetNotifyCallbackInterface(
 904                 [in] ISpNotifyCallback *callback,
 905                 [in] WPARAM wparam,
 906                 [in] LPARAM lparam);
 907     [local] HRESULT SetNotifyWin32Event(void);
 908     [local] HRESULT WaitForNotifyEvent([in] DWORD milliseconds);
 909     [local] HANDLE  GetNotifyEventHandle();
 910 }
 911
 912 [
 913     object,
 914     uuid(be7a9cce-5f9e-11d2-960f-00c04f8ee628),
 915     pointer_default(unique),
 916     local,
 917     restricted
 918 ]
 919 interface ISpEventSource : ISpNotifySource
 920 {
 921     HRESULT SetInterest([in] ULONGLONG event, [in] ULONGLONG queued);
 922     HRESULT GetEvents([in] ULONG count,
 923                     [out, size_is(count)] SPEVENT *array,
 924                     [out] ULONG *fetched);
 925     HRESULT GetInfo([out] SPEVENTSOURCEINFO *info);
 926 };
 927
 928 [
 929     object,
 930     uuid(f740a62f-7c15-489e-8234-940a33d9272d),
 931     pointer_default(unique),
 932     local,
 933     restricted
 934 ]
 935 interface ISpRecoContext : ISpEventSource
 936 {
 937     HRESULT GetRecognizer([out] ISpRecognizer **recognizer);
 938
 939     HRESULT CreateGrammar([in] ULONGLONG grammar, [out] ISpRecoGrammar **spgrammar);
 940
 941     HRESULT GetStatus([out] SPRECOCONTEXTSTATUS *status);
 942
 943     HRESULT GetMaxAlternates([in] ULONG *alternates);
 944     HRESULT SetMaxAlternates([in] ULONG alternates);
 945
 946     HRESULT SetAudioOptions([in] SPAUDIOOPTIONS options, [in] const GUID *format, [in] const WAVEFORMATEX *wave);
 947     HRESULT GetAudioOptions([in] SPAUDIOOPTIONS *options, [out] GUID *format, [out] WAVEFORMATEX **wave);
 948
 949     HRESULT DeserializeResult([in] const SPSERIALIZEDRESULT *serialized, [out] ISpRecoResult **result);
 950
 951     HRESULT Bookmark([in] SPBOOKMARKOPTIONS options, [in] ULONGLONG position, [in] LPARAM event);
 952
 953     HRESULT SetAdaptationData([in, string] const WCHAR *data, [in] const ULONG count);
 954
 955     HRESULT Pause(DWORD reserved);
 956     HRESULT Resume(DWORD reserved);
 957
 958     HRESULT SetVoice([in] ISpVoice *voice, [in] BOOL changes);
 959     HRESULT GetVoice([out] ISpVoice **voice);
 960     HRESULT SetVoicePurgeEvent([in] ULONGLONG interest);
 961     HRESULT GetVoicePurgeEvent([out] ULONGLONG *interest);
 962
 963     HRESULT SetContextState([in] SPCONTEXTSTATE state);
 964     HRESULT GetContextState([in] SPCONTEXTSTATE *state);
 965 };
 966
 967 [
 968     object,
 969     uuid(bed530be-2606-4f4d-a1c0-54c5cda5566f),
 970     pointer_default(unique),
 971     restricted
 972 ]
 973 interface ISpStreamFormat : IStream
 974 {
 975     HRESULT GetFormat(GUID *format, WAVEFORMATEX **wave);
 976 }
 977
 978 [
 979     object,
 980     uuid(5b4fb971-b115-4de1-ad97-e482e3bf6ee4),
 981     pointer_default(unique),
 982     local,
 983     restricted
 984 ]
 985 interface ISpProperties : IUnknown
 986 {
 987     HRESULT SetPropertyNum([in] const WCHAR *name, [in] long value);
 988     HRESULT GetPropertyNum([in] const WCHAR *name, [out] long *value);
 989     HRESULT SetPropertyString([in] const WCHAR *name, [in] const WCHAR *value);
 990     HRESULT GetPropertyString([in] const WCHAR *name, [out] WCHAR **value);
 991 };
 992
 993 [
 994     object,
 995     uuid(c2b5f241-daa0-4507-9e16-5a1eaa2b7a5c),
 996     pointer_default(unique),
 997     restricted
 998 ]
 999 interface ISpRecognizer : ISpProperties
1000 {
1001     HRESULT SetRecognizer([in] ISpObjectToken *recognizer);
1002     HRESULT GetRecognizer([out] ISpObjectToken **recognizer);
1003
1004     HRESULT SetInput([in] IUnknown *input, [in] BOOL changes);
1005     HRESULT GetInputObjectToken([out] ISpObjectToken **token);
1006
1007     HRESULT GetInputStream( [out] ISpStreamFormat **stream);
1008     HRESULT CreateRecoContext([out] ISpRecoContext **cxt);
1009
1010     HRESULT GetRecoProfile([out] ISpObjectToken **token);
1011     HRESULT SetRecoProfile([in] ISpObjectToken *token);
1012
1013     HRESULT IsSharedInstance(void);
1014
1015     HRESULT GetRecoState([out] SPRECOSTATE *state);
1016     HRESULT SetRecoState([in] SPRECOSTATE state);
1017
1018     HRESULT GetStatus([out] SPRECOGNIZERSTATUS *status);
1019     HRESULT GetFormat([in] SPSTREAMFORMATTYPE type, [out] GUID *format, [out] WAVEFORMATEX **wave);
1020
1021     [local] HRESULT IsUISupported([in] const WCHAR *type, [in] void *extra, [in] ULONG count,
1022                 [out] BOOL *supported);
1023     [local] HRESULT DisplayUI([in] HWND hwnd,[in] const WCHAR *title, [in] const WCHAR *type,
1024                 [in] void *extra, [in] ULONG cnt);
1025     HRESULT EmulateRecognition([in] ISpPhrase *phrase);
1026 };
1027
1028 [
1029     object,
1030     uuid(12e3cca9-7518-44c5-a5e7-ba5a79cb929e),
1031     pointer_default(unique),
1032     local,
1033     restricted
1034 ]
1035 interface ISpStream : ISpStreamFormat
1036 {
1037     HRESULT SetBaseStream([in] IStream *stream, [in] REFGUID format, [in] const WAVEFORMATEX *wave);
1038     HRESULT GetBaseStream([out] IStream **stream);
1039     HRESULT BindToFile([in] LPCWSTR filename, [in] SPFILEMODE mode, [in] const GUID *format,
1040                        const WAVEFORMATEX *wave, [in] ULONGLONG interest);
1041     HRESULT Close();
1042 }
1043
1044 [
1045     helpstring("Speech Object Library"),
1046     uuid(c866ca3a-32f7-11d2-9602-00c04f8ee628),
1047     version(5.4)
1048 ]
1049 library SpeechLib
1050 {
1051     importlib("stdole2.tlb");
1052
1053 #include "sapiaut.idl"
1054
1055     [
1056         uuid(a910187f-0c7a-45ac-92cc-59edafb77b53),
1057         helpstring("SpObjectTokenCategory Class"),
1058         progid("SAPI.SpObjectTokenCategory.1"),
1059         vi_progid("SAPI.SpObjectTokenCategory"),
1060         threading(both)
1061     ]
1062     coclass SpObjectTokenCategory
1063     {
1064         interface ISpObjectTokenCategory;
1065     }
1066
1067     [
1068         uuid(3bee4890-4fe9-4a37-8c1e-5e7e12791c1f)
1069     ]
1070     coclass SpSharedRecognizer
1071     {
1072         /*[default] interface ISpeechRecognizer;*/
1073         interface ISpRecognizer;
1074     };
1075
1076     [
1077         uuid(96749377-3391-11d2-9ee3-00c04f797396)
1078     ]
1079     coclass SpVoice
1080     {
1081         interface ISpVoice;
1082         [default] interface ISpeechVoice;
1083     };
1084
1085     [
1086         uuid(947812b3-2ae1-4644-ba86-9e90ded7ec91),
1087     ]
1088     coclass SpFileStream
1089     {
1090         interface ISpStream;
1091         [default] interface ISpeechFileStream;
1092     };
1093 }