Thinkai's Blog
Autohotkey|Python|php|aardio|VOIP|IT 爱好者
示例：
;朗读CMD传入的字符串
#SingleInstance Off
#NoTrayIcon
tts := ComObjCreate("SAPI.SpVoice")
tts.Voice := tts.GetVoices("Name=VW Hui").Item(0) ;把发音人设置成名称为VW Hui的
tts.Rate := 3 ;说话速度 -10到10
tts.Volume := 100 ;音量 0到100
txt = %1%
tts.Speak(txt) ;开始朗读
IDL接口文档：
// Generated .IDL file (by the OLE/COM Object Viewer)
//
// typelib filename: sapi.dll
[
uuid(C866CA3A-32F7-11D2-9602-00C04F8EE628),
version(5.4),
helpstring("Microsoft Speech Object Library")
]
library SpeechLib
{
// TLib : // TLib : OLE Automation : {00020430-0000-0000-C000-000000000046}
importlib("stdole2.tlb");
// Forward declare all types defined in this typelib
interface ISpeechDataKey;
interface ISpeechObjectToken;
interface ISpeechObjectTokenCategory;
interface ISpeechObjectTokens;
interface ISpeechAudioBufferInfo;
interface ISpeechAudioStatus;
interface ISpeechAudioFormat;
interface ISpeechWaveFormatEx;
interface ISpeechBaseStream;
interface ISpeechFileStream;
interface ISpeechMemoryStream;
interface ISpeechCustomStream;
interface ISpeechAudio;
interface ISpeechMMSysAudio;
interface ISpeechVoice;
interface ISpeechVoiceStatus;
dispinterface _ISpeechVoiceEvents;
interface ISpeechRecognizer;
interface ISpeechRecognizerStatus;
interface ISpeechRecoContext;
interface ISpeechRecoGrammar;
interface ISpeechGrammarRules;
interface ISpeechGrammarRule;
interface ISpeechGrammarRuleState;
interface ISpeechGrammarRuleStateTransitions;
interface ISpeechGrammarRuleStateTransition;
interface ISpeechTextSelectionInformation;
interface ISpeechRecoResult;
interface ISpeechRecoResultTimes;
interface ISpeechPhraseInfo;
interface ISpeechPhraseRule;
interface ISpeechPhraseRules;
interface ISpeechPhraseProperties;
interface ISpeechPhraseProperty;
interface ISpeechPhraseElements;
interface ISpeechPhraseElement;
interface ISpeechPhraseReplacements;
interface ISpeechPhraseReplacement;
interface ISpeechPhraseAlternates;
interface ISpeechPhraseAlternate;
dispinterface _ISpeechRecoContextEvents;
interface ISpeechRecoResult2;
interface ISpeechLexicon;
interface ISpeechLexiconWords;
interface ISpeechLexiconWord;
interface ISpeechLexiconPronunciations;
interface ISpeechLexiconPronunciation;
interface ISpeechXMLRecoResult;
interface ISpeechRecoResultDispatch;
interface ISpeechPhraseInfoBuilder;
interface ISpeechPhoneConverter;
interface ISpNotifyTranslator;
interface ISpNotifySink;
interface ISpObjectTokenCategory;
interface ISpDataKey;
interface IEnumSpObjectTokens;
interface ISpObjectToken;
interface ISpResourceManager;
interface IServiceProvider;
interface ISpStreamFormatConverter;
interface ISpStreamFormat;
interface IStream;
interface ISequentialStream;
interface ISpEventSource;
interface ISpNotifySource;
interface ISpEventSink;
interface ISpObjectWithToken;
interface ISpMMSysAudio;
interface ISpAudio;
interface ISpStream;
interface ISpVoice;
interface ISpPhoneticAlphabetSelection;
interface ISpRecoContext;
interface ISpRecognizer;
interface ISpProperties;
interface ISpPhrase;
interface ISpRecoGrammar;
interface ISpGrammarBuilder;
interface ISpRecoResult;
interface ISpPhraseAlt;
interface ISpRecoContext2;
interface ISpRecognizer2;
interface ISpRecognizer3;
interface ISpRecoCategory;
interface ISpSerializeState;
interface ISpLexicon;
interface ISpShortcut;
interface ISpPhoneConverter;
interface ISpPhoneticAlphabetConverter;
interface ISpXMLRecoResult;
interface ISpRecoGrammar2;
interface ISpeechResourceLoader;
interface IInternetSecurityManager;
interface IInternetSecurityMgrSite;
interface IEnumString;
[
odl,
uuid(CE17C09B-4EFA-44D5-A4C9-59D9585AB0CD),
helpstring("ISpeechDataKey Interface"),
dual,
oleautomation
]
interface ISpeechDataKey : IDispatch {
[id(0x00000001), helpstring("SetBinaryValue")]
HRESULT SetBinaryValue(
[in] BSTR ValueName,
[in] VARIANT Value);
[id(0x00000002), helpstring("GetBinaryValue")]
HRESULT GetBinaryValue(
[in] BSTR ValueName,
[out, retval] VARIANT* Value);
[id(0x00000003), helpstring("SetStringValue")]
HRESULT SetStringValue(
[in] BSTR ValueName,
[in] BSTR Value);
[id(0x00000004), helpstring("GetStringValue")]
HRESULT GetStringValue(
[in] BSTR ValueName,
[out, retval] BSTR* Value);
[id(0x00000005), helpstring("SetLongValue")]
HRESULT SetLongValue(
[in] BSTR ValueName,
[in] long Value);
[id(0x00000006), helpstring("GetlongValue")]
HRESULT GetLongValue(
[in] BSTR ValueName,
[out, retval] long* Value);
[id(0x00000007), helpstring("OpenKey")]
HRESULT OpenKey(
[in] BSTR SubKeyName,
[out, retval] ISpeechDataKey** SubKey);
[id(0x00000008), helpstring("CreateKey")]
HRESULT CreateKey(
[in] BSTR SubKeyName,
[out, retval] ISpeechDataKey** SubKey);
[id(0x00000009), helpstring("DeleteKey")]
HRESULT DeleteKey([in] BSTR SubKeyName);
[id(0x0000000a), helpstring("DeleteValue")]
HRESULT DeleteValue([in] BSTR ValueName);
[id(0x0000000b), helpstring("EnumKeys")]
HRESULT EnumKeys(
[in] long Index,
[out, retval] BSTR* SubKeyName);
[id(0x0000000c), helpstring("EnumValues")]
HRESULT EnumValues(
[in] long Index,
[out, retval] BSTR* ValueName);
};
[
odl,
uuid(C74A3ADC-B727-4500-A84A-B526721C8B8C),
helpstring("ISpeechObjectToken Interface"),
dual,
oleautomation
]
interface ISpeechObjectToken : IDispatch {
[id(0x00000001), propget, helpstring("Id")]
HRESULT Id([out, retval] BSTR* ObjectId);
[id(0x00000002), propget, hidden, helpstring("DataKey")]
HRESULT DataKey([out, retval] ISpeechDataKey** DataKey);
[id(0x00000003), propget, helpstring("Category")]
HRESULT Category([out, retval] ISpeechObjectTokenCategory** Category);
[id(0x00000004), helpstring("GetDescription")]
HRESULT GetDescription(
[in, optional, defaultvalue(0)] long Locale,
[out, retval] BSTR* Description);
[id(0x00000005), hidden, helpstring("SetId")]
HRESULT SetId(
[in] BSTR Id,
[in, optional, defaultvalue("")] BSTR CategoryID,
[in, optional, defaultvalue(0)] VARIANT_BOOL CreateIfNotExist);
[id(0x00000006), helpstring("GetAttribute")]
HRESULT GetAttribute(
[in] BSTR AttributeName,
[out, retval] BSTR* AttributeValue);
[id(0x00000007), helpstring("CreateInstance")]
HRESULT CreateInstance(
[in, optional, defaultvalue(<unprintable IUnknown*>)] IUnknown* pUnkOuter,
[in, optional, defaultvalue(23)] SpeechTokenContext ClsContext,
[out, retval] IUnknown** Object);
[id(0x00000008), hidden, helpstring("Remove")]
HRESULT Remove([in] BSTR ObjectStorageCLSID);
[id(0x00000009), hidden, helpstring("GetStorageFileName")]
HRESULT GetStorageFileName(
[in] BSTR ObjectStorageCLSID,
[in] BSTR KeyName,
[in] BSTR FileName,
[in] SpeechTokenShellFolder Folder,
[out, retval] BSTR* FilePath);
[id(0x0000000a), hidden, helpstring("RemoveStorageFileName")]
HRESULT RemoveStorageFileName(
[in] BSTR ObjectStorageCLSID,
[in] BSTR KeyName,
[in] VARIANT_BOOL DeleteFile);
[id(0x0000000b), hidden, helpstring("IsUISupported")]
HRESULT IsUISupported(
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData,
[in, optional, defaultvalue(<unprintable IUnknown*>)] IUnknown* Object,
[out, retval] VARIANT_BOOL* Supported);
[id(0x0000000c), hidden, helpstring("DisplayUI")]
HRESULT DisplayUI(
[in] long hWnd,
[in] BSTR Title,
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData,
[in, optional, defaultvalue(<unprintable IUnknown*>)] IUnknown* Object);
[id(0x0000000d), helpstring("MatchesAttributes")]
HRESULT MatchesAttributes(
[in] BSTR Attributes,
[out, retval] VARIANT_BOOL* Matches);
};
[
odl,
uuid(CA7EAC50-2D01-4145-86D4-5AE7D70F4469),
helpstring("ISpeechObjectTokenCategory Interface"),
dual,
oleautomation
]
interface ISpeechObjectTokenCategory : IDispatch {
[id(0x00000001), propget, helpstring("Id")]
HRESULT Id([out, retval] BSTR* Id);
[id(0x00000002), propput, helpstring("Default")]
HRESULT Default([in] BSTR TokenId);
[id(0x00000002), propget, helpstring("Default")]
HRESULT Default([out, retval] BSTR* TokenId);
[id(0x00000003), helpstring("SetId")]
HRESULT SetId(
[in] BSTR Id,
[in, optional, defaultvalue(0)] VARIANT_BOOL CreateIfNotExist);
[id(0x00000004), hidden, helpstring("GetDataKey")]
HRESULT GetDataKey(
[in, optional, defaultvalue(0)] SpeechDataKeyLocation Location,
[out, retval] ISpeechDataKey** DataKey);
[id(0x00000005), helpstring("EnumerateTokens")]
HRESULT EnumerateTokens(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** Tokens);
};
typedef enum {
SDKLDefaultLocation = 0,
SDKLCurrentUser = 1,
SDKLLocalMachine = 2,
SDKLCurrentConfig = 5
} SpeechDataKeyLocation;
[
odl,
uuid(9285B776-2E7B-4BC0-B53E-580EB6FA967F),
helpstring("ISpeechObjectTokens Interface"),
dual,
oleautomation
]
interface ISpeechObjectTokens : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechObjectToken** Token);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the tokens")]
HRESULT _NewEnum([out, retval] IUnknown** ppEnumVARIANT);
};
typedef enum {
STCInprocServer = 1,
STCInprocHandler = 2,
STCLocalServer = 4,
STCRemoteServer = 16,
STCAll = 23
} SpeechTokenContext;
typedef enum {
STSF_AppData = 26,
STSF_LocalAppData = 28,
STSF_CommonAppData = 35,
STSF_FlagCreate = 32768
} SpeechTokenShellFolder;
[
odl,
uuid(11B103D8-1142-4EDF-A093-82FB3915F8CC),
helpstring("ISpeechAudioBufferInfo Interface"),
dual,
oleautomation
]
interface ISpeechAudioBufferInfo : IDispatch {
[id(0x00000001), propget, helpstring("MinNotification")]
HRESULT MinNotification([out, retval] long* MinNotification);
[id(0x00000001), propput, helpstring("MinNotification")]
HRESULT MinNotification([in] long MinNotification);
[id(0x00000002), propget, helpstring("BufferSize")]
HRESULT BufferSize([out, retval] long* BufferSize);
[id(0x00000002), propput, helpstring("BufferSize")]
HRESULT BufferSize([in] long BufferSize);
[id(0x00000003), propget, helpstring("EventBias")]
HRESULT EventBias([out, retval] long* EventBias);
[id(0x00000003), propput, helpstring("EventBias")]
HRESULT EventBias([in] long EventBias);
};
[
odl,
uuid(C62D9C91-7458-47F6-862D-1EF86FB0B278),
helpstring("ISpeechAudioStatus Interface"),
dual,
oleautomation
]
interface ISpeechAudioStatus : IDispatch {
[id(0x00000001), propget, helpstring("FreeBufferSpace")]
HRESULT FreeBufferSpace([out, retval] long* FreeBufferSpace);
[id(0x00000002), propget, helpstring("NonBlockingIO")]
HRESULT NonBlockingIO([out, retval] long* NonBlockingIO);
[id(0x00000003), propget, helpstring("State")]
HRESULT State([out, retval] SpeechAudioState* State);
[id(0x00000004), propget, helpstring("CurrentSeekPosition")]
HRESULT CurrentSeekPosition([out, retval] VARIANT* CurrentSeekPosition);
[id(0x00000005), propget, helpstring("CurrentDevicePosition")]
HRESULT CurrentDevicePosition([out, retval] VARIANT* CurrentDevicePosition);
};
typedef enum {
SASClosed = 0,
SASStop = 1,
SASPause = 2,
SASRun = 3
} SpeechAudioState;
[
odl,
uuid(E6E9C590-3E18-40E3-8299-061F98BDE7C7),
helpstring("ISpeechAudioFormat Interface"),
dual,
oleautomation
]
interface ISpeechAudioFormat : IDispatch {
[id(0x00000001), propget, helpstring("Type")]
HRESULT Type([out, retval] SpeechAudioFormatType* AudioFormat);
[id(0x00000001), propput, helpstring("Type")]
HRESULT Type([in] SpeechAudioFormatType AudioFormat);
[id(0x00000002), propget, hidden, helpstring("Guid")]
HRESULT Guid([out, retval] BSTR* Guid);
[id(0x00000002), propput, hidden, helpstring("Guid")]
HRESULT Guid([in] BSTR Guid);
[id(0x00000003), hidden, helpstring("GetWaveFormatEx")]
HRESULT GetWaveFormatEx([out, retval] ISpeechWaveFormatEx** SpeechWaveFormatEx);
[id(0x00000004), hidden, helpstring("SetWaveFormatEx")]
HRESULT SetWaveFormatEx([in] ISpeechWaveFormatEx* SpeechWaveFormatEx);
};
typedef enum {
SAFTDefault = 0xffffffff,
SAFTNoAssignedFormat = 0,
SAFTText = 1,
SAFTNonStandardFormat = 2,
SAFTExtendedAudioFormat = 3,
SAFT8kHz8BitMono = 4,
SAFT8kHz8BitStereo = 5,
SAFT8kHz16BitMono = 6,
SAFT8kHz16BitStereo = 7,
SAFT11kHz8BitMono = 8,
SAFT11kHz8BitStereo = 9,
SAFT11kHz16BitMono = 10,
SAFT11kHz16BitStereo = 11,
SAFT12kHz8BitMono = 12,
SAFT12kHz8BitStereo = 13,
SAFT12kHz16BitMono = 14,
SAFT12kHz16BitStereo = 15,
SAFT16kHz8BitMono = 16,
SAFT16kHz8BitStereo = 17,
SAFT16kHz16BitMono = 18,
SAFT16kHz16BitStereo = 19,
SAFT22kHz8BitMono = 20,
SAFT22kHz8BitStereo = 21,
SAFT22kHz16BitMono = 22,
SAFT22kHz16BitStereo = 23,
SAFT24kHz8BitMono = 24,
SAFT24kHz8BitStereo = 25,
SAFT24kHz16BitMono = 26,
SAFT24kHz16BitStereo = 27,
SAFT32kHz8BitMono = 28,
SAFT32kHz8BitStereo = 29,
SAFT32kHz16BitMono = 30,
SAFT32kHz16BitStereo = 31,
SAFT44kHz8BitMono = 32,
SAFT44kHz8BitStereo = 33,
SAFT44kHz16BitMono = 34,
SAFT44kHz16BitStereo = 35,
SAFT48kHz8BitMono = 36,
SAFT48kHz8BitStereo = 37,
SAFT48kHz16BitMono = 38,
SAFT48kHz16BitStereo = 39,
SAFTTrueSpeech_8kHz1BitMono = 40,
SAFTCCITT_ALaw_8kHzMono = 41,
SAFTCCITT_ALaw_8kHzStereo = 42,
SAFTCCITT_ALaw_11kHzMono = 43,
SAFTCCITT_ALaw_11kHzStereo = 44,
SAFTCCITT_ALaw_22kHzMono = 45,
SAFTCCITT_ALaw_22kHzStereo = 46,
SAFTCCITT_ALaw_44kHzMono = 47,
SAFTCCITT_ALaw_44kHzStereo = 48,
SAFTCCITT_uLaw_8kHzMono = 49,
SAFTCCITT_uLaw_8kHzStereo = 50,
SAFTCCITT_uLaw_11kHzMono = 51,
SAFTCCITT_uLaw_11kHzStereo = 52,
SAFTCCITT_uLaw_22kHzMono = 53,
SAFTCCITT_uLaw_22kHzStereo = 54,
SAFTCCITT_uLaw_44kHzMono = 55,
SAFTCCITT_uLaw_44kHzStereo = 56,
SAFTADPCM_8kHzMono = 57,
SAFTADPCM_8kHzStereo = 58,
SAFTADPCM_11kHzMono = 59,
SAFTADPCM_11kHzStereo = 60,
SAFTADPCM_22kHzMono = 61,
SAFTADPCM_22kHzStereo = 62,
SAFTADPCM_44kHzMono = 63,
SAFTADPCM_44kHzStereo = 64,
SAFTGSM610_8kHzMono = 65,
SAFTGSM610_11kHzMono = 66,
SAFTGSM610_22kHzMono = 67,
SAFTGSM610_44kHzMono = 68
} SpeechAudioFormatType;
[
odl,
uuid(7A1EF0D5-1581-4741-88E4-209A49F11A10),
helpstring("ISpeechWaveFormatEx Interface"),
dual,
oleautomation
]
interface ISpeechWaveFormatEx : IDispatch {
[id(0x00000001), propget, helpstring("FormatTag")]
HRESULT FormatTag([out, retval] short* FormatTag);
[id(0x00000001), propput, helpstring("FormatTag")]
HRESULT FormatTag([in] short FormatTag);
[id(0x00000002), propget, helpstring("Channels")]
HRESULT Channels([out, retval] short* Channels);
[id(0x00000002), propput, helpstring("Channels")]
HRESULT Channels([in] short Channels);
[id(0x00000003), propget, helpstring("SamplesPerSec")]
HRESULT SamplesPerSec([out, retval] long* SamplesPerSec);
[id(0x00000003), propput, helpstring("SamplesPerSec")]
HRESULT SamplesPerSec([in] long SamplesPerSec);
[id(0x00000004), propget, helpstring("AvgBytesPerSec")]
HRESULT AvgBytesPerSec([out, retval] long* AvgBytesPerSec);
[id(0x00000004), propput, helpstring("AvgBytesPerSec")]
HRESULT AvgBytesPerSec([in] long AvgBytesPerSec);
[id(0x00000005), propget, helpstring("BlockAlign")]
HRESULT BlockAlign([out, retval] short* BlockAlign);
[id(0x00000005), propput, helpstring("BlockAlign")]
HRESULT BlockAlign([in] short BlockAlign);
[id(0x00000006), propget, helpstring("BitsPerSample")]
HRESULT BitsPerSample([out, retval] short* BitsPerSample);
[id(0x00000006), propput, helpstring("BitsPerSample")]
HRESULT BitsPerSample([in] short BitsPerSample);
[id(0x00000007), propget, helpstring("ExtraData")]
HRESULT ExtraData([out, retval] VARIANT* ExtraData);
[id(0x00000007), propput, helpstring("ExtraData")]
HRESULT ExtraData([in] VARIANT ExtraData);
};
[
odl,
uuid(6450336F-7D49-4CED-8097-49D6DEE37294),
helpstring("ISpeechBaseStream Interface"),
dual,
oleautomation
]
interface ISpeechBaseStream : IDispatch {
[id(0x00000001), propget, helpstring("Format")]
HRESULT Format([out, retval] ISpeechAudioFormat** AudioFormat);
[id(0x00000001), propputref, helpstring("Format")]
HRESULT Format([in] ISpeechAudioFormat* AudioFormat);
[id(0x00000002), helpstring("Read")]
HRESULT Read(
[out] VARIANT* Buffer,
[in] long NumberOfBytes,
[out, retval] long* BytesRead);
[id(0x00000003), helpstring("Write")]
HRESULT Write(
[in] VARIANT Buffer,
[out, retval] long* BytesWritten);
[id(0x00000004), helpstring("Seek")]
HRESULT Seek(
[in] VARIANT Position,
[in, optional, defaultvalue(0)] SpeechStreamSeekPositionType Origin,
[out, retval] VARIANT* NewPosition);
};
typedef enum {
SSSPTRelativeToStart = 0,
SSSPTRelativeToCurrentPosition = 1,
SSSPTRelativeToEnd = 2
} SpeechStreamSeekPositionType;
[
odl,
uuid(AF67F125-AB39-4E93-B4A2-CC2E66E182A7),
helpstring("ISpeechFileStream Interface"),
dual,
oleautomation
]
interface ISpeechFileStream : ISpeechBaseStream {
[id(0x00000064), helpstring("Open")]
HRESULT Open(
[in] BSTR FileName,
[in, optional, defaultvalue(0)] SpeechStreamFileMode FileMode,
[in, optional, defaultvalue(0)] VARIANT_BOOL DoEvents);
[id(0x00000065), helpstring("Close")]
HRESULT Close();
};
typedef enum {
SSFMOpenForRead = 0,
SSFMOpenReadWrite = 1,
SSFMCreate = 2,
SSFMCreateForWrite = 3
} SpeechStreamFileMode;
[
odl,
uuid(EEB14B68-808B-4ABE-A5EA-B51DA7588008),
helpstring("ISpeechMemoryStream Interface"),
dual,
oleautomation
]
interface ISpeechMemoryStream : ISpeechBaseStream {
[id(0x00000064), helpstring("SetData")]
HRESULT SetData([in] VARIANT Data);
[id(0x00000065), helpstring("GetData")]
HRESULT GetData([out, retval] VARIANT* pData);
};
[
odl,
uuid(1A9E9F4F-104F-4DB8-A115-EFD7FD0C97AE),
helpstring("ISpeechCustomStream Interface"),
dual,
oleautomation
]
interface ISpeechCustomStream : ISpeechBaseStream {
[id(0x00000064), propget, helpstring("BaseStream")]
HRESULT BaseStream([out, retval] IUnknown** ppUnkStream);
[id(0x00000064), propputref, helpstring("BaseStream")]
HRESULT BaseStream([in] IUnknown* ppUnkStream);
};
[
odl,
uuid(CFF8E175-019E-11D3-A08E-00C04F8EF9B5),
helpstring("ISpeechAudio Interface"),
dual,
oleautomation
]
interface ISpeechAudio : ISpeechBaseStream {
[id(0x000000c8), propget, helpstring("Status")]
HRESULT Status([out, retval] ISpeechAudioStatus** Status);
[id(0x000000c9), propget, helpstring("BufferInfo")]
HRESULT BufferInfo([out, retval] ISpeechAudioBufferInfo** BufferInfo);
[id(0x000000ca), propget, helpstring("DefaultFormat")]
HRESULT DefaultFormat([out, retval] ISpeechAudioFormat** StreamFormat);
[id(0x000000cb), propget, helpstring("Volume")]
HRESULT Volume([out, retval] long* Volume);
[id(0x000000cb), propput, helpstring("Volume")]
HRESULT Volume([in] long Volume);
[id(0x000000cc), propget, helpstring("BufferNotifySize")]
HRESULT BufferNotifySize([out, retval] long* BufferNotifySize);
[id(0x000000cc), propput, helpstring("BufferNotifySize")]
HRESULT BufferNotifySize([in] long BufferNotifySize);
[id(0x000000cd), propget, hidden, helpstring("EventHandle")]
HRESULT EventHandle([out, retval] long* EventHandle);
[id(0x000000ce), hidden, helpstring("SetState")]
HRESULT SetState([in] SpeechAudioState State);
};
[
odl,
uuid(3C76AF6D-1FD7-4831-81D1-3B71D5A13C44),
helpstring("ISpeechMMSysAudio Interface"),
dual,
oleautomation
]
interface ISpeechMMSysAudio : ISpeechAudio {
[id(0x0000012c), propget, helpstring("DeviceId")]
HRESULT DeviceId([out, retval] long* DeviceId);
[id(0x0000012c), propput, helpstring("DeviceId")]
HRESULT DeviceId([in] long DeviceId);
[id(0x0000012d), propget, helpstring("LineId")]
HRESULT LineId([out, retval] long* LineId);
[id(0x0000012d), propput, helpstring("LineId")]
HRESULT LineId([in] long LineId);
[id(0x0000012e), propget, hidden, helpstring("MMHandle")]
HRESULT MMHandle([out, retval] long* Handle);
};
[
odl,
uuid(269316D8-57BD-11D2-9EEE-00C04F797396),
helpstring("ISpeechVoice Interface"),
dual,
oleautomation
]
interface ISpeechVoice : IDispatch {
[id(0x00000001), propget, helpstring("Status")]
HRESULT Status([out, retval] ISpeechVoiceStatus** Status);
[id(0x00000002), propget, helpstring("Voice")]
HRESULT Voice([out, retval] ISpeechObjectToken** Voice);
[id(0x00000002), propputref, helpstring("Voice")]
HRESULT Voice([in] ISpeechObjectToken* Voice);
[id(0x00000003), propget, helpstring("Gets the audio output object")]
HRESULT AudioOutput([out, retval] ISpeechObjectToken** AudioOutput);
[id(0x00000003), propputref, helpstring("Gets the audio output object")]
HRESULT AudioOutput([in] ISpeechObjectToken* AudioOutput);
[id(0x00000004), propget, helpstring("Gets the audio output stream")]
HRESULT AudioOutputStream([out, retval] ISpeechBaseStream** AudioOutputStream);
[id(0x00000004), propputref, helpstring("Gets the audio output stream")]
HRESULT AudioOutputStream([in] ISpeechBaseStream* AudioOutputStream);
[id(0x00000005), propget, helpstring("Rate")]
HRESULT Rate([out, retval] long* Rate);
[id(0x00000005), propput, helpstring("Rate")]
HRESULT Rate([in] long Rate);
[id(0x00000006), propget, helpstring("Volume")]
HRESULT Volume([out, retval] long* Volume);
[id(0x00000006), propput, helpstring("Volume")]
HRESULT Volume([in] long Volume);
[id(0x00000007), propput, hidden, helpstring("AllowAudioOutputFormatChangesOnNextSet")]
HRESULT AllowAudioOutputFormatChangesOnNextSet([in] VARIANT_BOOL Allow);
[id(0x00000007), propget, hidden, helpstring("AllowAudioOutputFormatChangesOnNextSet")]
HRESULT AllowAudioOutputFormatChangesOnNextSet([out, retval] VARIANT_BOOL* Allow);
[id(0x00000008), propget, helpstring("EventInterests")]
HRESULT EventInterests([out, retval] SpeechVoiceEvents* EventInterestFlags);
[id(0x00000008), propput, helpstring("EventInterests")]
HRESULT EventInterests([in] SpeechVoiceEvents EventInterestFlags);
[id(0x00000009), propput, helpstring("Priority")]
HRESULT Priority([in] SpeechVoicePriority Priority);
[id(0x00000009), propget, helpstring("Priority")]
HRESULT Priority([out, retval] SpeechVoicePriority* Priority);
[id(0x0000000a), propput, helpstring("AlertBoundary")]
HRESULT AlertBoundary([in] SpeechVoiceEvents Boundary);
[id(0x0000000a), propget, helpstring("AlertBoundary")]
HRESULT AlertBoundary([out, retval] SpeechVoiceEvents* Boundary);
[id(0x0000000b), propput, helpstring("SyncSpeakTimeout")]
HRESULT SynchronousSpeakTimeout([in] long msTimeout);
[id(0x0000000b), propget, helpstring("SyncSpeakTimeout")]
HRESULT SynchronousSpeakTimeout([out, retval] long* msTimeout);
[id(0x0000000c), helpstring("Speak")]
HRESULT Speak(
[in] BSTR Text,
[in, optional, defaultvalue(0)] SpeechVoiceSpeakFlags Flags,
[out, retval] long* StreamNumber);
[id(0x0000000d), helpstring("SpeakStream")]
HRESULT SpeakStream(
[in] ISpeechBaseStream* Stream,
[in, optional, defaultvalue(0)] SpeechVoiceSpeakFlags Flags,
[out, retval] long* StreamNumber);
[id(0x0000000e), helpstring("Pauses the voices rendering.")]
HRESULT Pause();
[id(0x0000000f), helpstring("Resumes the voices rendering.")]
HRESULT Resume();
[id(0x00000010), helpstring("Skips rendering the specified number of items.")]
HRESULT Skip(
[in] BSTR Type,
[in] long NumItems,
[out, retval] long* NumSkipped);
[id(0x00000011), helpstring("GetVoices")]
HRESULT GetVoices(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** ObjectTokens);
[id(0x00000012), helpstring("GetAudioOutputs")]
HRESULT GetAudioOutputs(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** ObjectTokens);
[id(0x00000013), helpstring("WaitUntilDone")]
HRESULT WaitUntilDone(
[in] long msTimeout,
[out, retval] VARIANT_BOOL* Done);
[id(0x00000014), hidden, helpstring("SpeakCompleteEvent")]
HRESULT SpeakCompleteEvent([out, retval] long* Handle);
[id(0x00000015), helpstring("IsUISupported")]
HRESULT IsUISupported(
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData,
[out, retval] VARIANT_BOOL* Supported);
[id(0x00000016), helpstring("DisplayUI")]
HRESULT DisplayUI(
[in] long hWndParent,
[in] BSTR Title,
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData);
};
[
odl,
uuid(8BE47B07-57F6-11D2-9EEE-00C04F797396),
helpstring("ISpeechVoiceStatus Interface"),
dual,
oleautomation
]
interface ISpeechVoiceStatus : IDispatch {
[id(0x00000001), propget, helpstring("CurrentStreamNumber")]
HRESULT CurrentStreamNumber([out, retval] long* StreamNumber);
[id(0x00000002), propget, helpstring("LastStreamNumberQueued")]
HRESULT LastStreamNumberQueued([out, retval] long* StreamNumber);
[id(0x00000003), propget, helpstring("LastHResult")]
HRESULT LastHResult([out, retval] long* HResult);
[id(0x00000004), propget, helpstring("RunningState")]
HRESULT RunningState([out, retval] SpeechRunState* State);
[id(0x00000005), propget, helpstring("InputWordPosition")]
HRESULT InputWordPosition([out, retval] long* Position);
[id(0x00000006), propget, helpstring("InputWordLength")]
HRESULT InputWordLength([out, retval] long* Length);
[id(0x00000007), propget, helpstring("InputSentencePosition")]
HRESULT InputSentencePosition([out, retval] long* Position);
[id(0x00000008), propget, helpstring("InputSentenceLength")]
HRESULT InputSentenceLength([out, retval] long* Length);
[id(0x00000009), propget, helpstring("LastBookmark")]
HRESULT LastBookmark([out, retval] BSTR* Bookmark);
[id(0x0000000a), propget, hidden, helpstring("LastBookmarkId")]
HRESULT LastBookmarkId([out, retval] long* BookmarkId);
[id(0x0000000b), propget, helpstring("PhonemeId")]
HRESULT PhonemeId([out, retval] short* PhoneId);
[id(0x0000000c), propget, helpstring("VisemeId")]
HRESULT VisemeId([out, retval] short* VisemeId);
};
typedef enum {
SRSEDone = 1,
SRSEIsSpeaking = 2
} SpeechRunState;
typedef enum {
SVEStartInputStream = 2,
SVEEndInputStream = 4,
SVEVoiceChange = 8,
SVEBookmark = 16,
SVEWordBoundary = 32,
SVEPhoneme = 64,
SVESentenceBoundary = 128,
SVEViseme = 256,
SVEAudioLevel = 512,
SVEPrivate = 32768,
SVEAllEvents = 33790
} SpeechVoiceEvents;
typedef enum {
SVPNormal = 0,
SVPAlert = 1,
SVPOver = 2
} SpeechVoicePriority;
typedef enum {
SVSFDefault = 0,
SVSFlagsAsync = 1,
SVSFPurgeBeforeSpeak = 2,
SVSFIsFilename = 4,
SVSFIsXML = 8,
SVSFIsNotXML = 16,
SVSFPersistXML = 32,
SVSFNLPSpeakPunc = 64,
SVSFParseSapi = 128,
SVSFParseSsml = 256,
SVSFParseAutodetect = 0,
SVSFNLPMask = 64,
SVSFParseMask = 384,
SVSFVoiceMask = 511,
SVSFUnusedFlags = 0xfffffe00
} SpeechVoiceSpeakFlags;
[
uuid(A372ACD1-3BEF-4BBD-8FFB-CB3E2B416AF8)
]
dispinterface _ISpeechVoiceEvents {
properties:
methods:
[id(0x00000001), helpstring("StartStream")]
void StartStream(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000002), helpstring("EndStream")]
void EndStream(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000003), helpstring("VoiceChange")]
void VoiceChange(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] ISpeechObjectToken* VoiceObjectToken);
[id(0x00000004), helpstring("Bookmark")]
void Bookmark(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] BSTR Bookmark,
[in] long BookmarkId);
[id(0x00000005), helpstring("Word")]
void Word(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long CharacterPosition,
[in] long Length);
[id(0x00000007), helpstring("Sentence")]
void Sentence(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long CharacterPosition,
[in] long Length);
[id(0x00000006), helpstring("Phoneme")]
void Phoneme(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long Duration,
[in] short NextPhoneId,
[in] SpeechVisemeFeature Feature,
[in] short CurrentPhoneId);
[id(0x00000008), helpstring("Viseme")]
void Viseme(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long Duration,
[in] SpeechVisemeType NextVisemeId,
[in] SpeechVisemeFeature Feature,
[in] SpeechVisemeType CurrentVisemeId);
[id(0x00000009), helpstring("AudioLevel")]
void AudioLevel(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long AudioLevel);
[id(0x0000000a), helpstring("EnginePrivate")]
void EnginePrivate(
[in] long StreamNumber,
[in] long StreamPosition,
[in] VARIANT EngineData);
};
typedef enum {
SVF_None = 0,
SVF_Stressed = 1,
SVF_Emphasis = 2
} SpeechVisemeFeature;
typedef enum {
SVP_0 = 0,
SVP_1 = 1,
SVP_2 = 2,
SVP_3 = 3,
SVP_4 = 4,
SVP_5 = 5,
SVP_6 = 6,
SVP_7 = 7,
SVP_8 = 8,
SVP_9 = 9,
SVP_10 = 10,
SVP_11 = 11,
SVP_12 = 12,
SVP_13 = 13,
SVP_14 = 14,
SVP_15 = 15,
SVP_16 = 16,
SVP_17 = 17,
SVP_18 = 18,
SVP_19 = 19,
SVP_20 = 20,
SVP_21 = 21
} SpeechVisemeType;
[
odl,
uuid(2D5F1C0C-BD75-4B08-9478-3B11FEA2586C),
helpstring("ISpeechRecognizer Interface"),
dual,
oleautomation
]
interface ISpeechRecognizer : IDispatch {
[id(0x00000001), propputref, helpstring("Recognizer")]
HRESULT Recognizer([in] ISpeechObjectToken* Recognizer);
[id(0x00000001), propget, helpstring("Recognizer")]
HRESULT Recognizer([out, retval] ISpeechObjectToken** Recognizer);
[id(0x00000002), propput, hidden, helpstring("AllowAudioInputFormatChangesOnNextSet")]
HRESULT AllowAudioInputFormatChangesOnNextSet([in] VARIANT_BOOL Allow);
[id(0x00000002), propget, hidden, helpstring("AllowAudioInputFormatChangesOnNextSet")]
HRESULT AllowAudioInputFormatChangesOnNextSet([out, retval] VARIANT_BOOL* Allow);
[id(0x00000003), propputref, helpstring("AudioInput")]
HRESULT AudioInput([in, optional, defaultvalue(0)] ISpeechObjectToken* AudioInput);
[id(0x00000003), propget, helpstring("AudioInput")]
HRESULT AudioInput([out, retval] ISpeechObjectToken** AudioInput);
[id(0x00000004), propputref, helpstring("AudioInputStream")]
HRESULT AudioInputStream([in, optional, defaultvalue(0)] ISpeechBaseStream* AudioInputStream);
[id(0x00000004), propget, helpstring("AudioInputStream")]
HRESULT AudioInputStream([out, retval] ISpeechBaseStream** AudioInputStream);
[id(0x00000005), propget, helpstring("IsShared")]
HRESULT IsShared([out, retval] VARIANT_BOOL* Shared);
[id(0x00000006), propput, helpstring("State")]
HRESULT State([in] SpeechRecognizerState State);
[id(0x00000006), propget, helpstring("State")]
HRESULT State([out, retval] SpeechRecognizerState* State);
[id(0x00000007), propget, helpstring("Status")]
HRESULT Status([out, retval] ISpeechRecognizerStatus** Status);
[id(0x00000008), propputref, helpstring("Profile")]
HRESULT Profile([in, optional, defaultvalue(0)] ISpeechObjectToken* Profile);
[id(0x00000008), propget, helpstring("Profile")]
HRESULT Profile([out, retval] ISpeechObjectToken** Profile);
[id(0x00000009), helpstring("EmulateRecognition")]
HRESULT EmulateRecognition(
[in] VARIANT TextElements,
[in, optional, defaultvalue("")] VARIANT* ElementDisplayAttributes,
[in, optional, defaultvalue(0)] long LanguageId);
[id(0x0000000a), helpstring("CreateRecoContext")]
HRESULT CreateRecoContext([out, retval] ISpeechRecoContext** NewContext);
[id(0x0000000b), helpstring("GetFormat")]
HRESULT GetFormat(
[in] SpeechFormatType Type,
[out, retval] ISpeechAudioFormat** Format);
[id(0x0000000c), hidden, helpstring("SetPropertyNumber")]
HRESULT SetPropertyNumber(
[in] BSTR Name,
[in] long Value,
[out, retval] VARIANT_BOOL* Supported);
[id(0x0000000d), hidden, helpstring("GetPropertyNumber")]
HRESULT GetPropertyNumber(
[in] BSTR Name,
[in, out] long* Value,
[out, retval] VARIANT_BOOL* Supported);
[id(0x0000000e), hidden, helpstring("SetPropertyString")]
HRESULT SetPropertyString(
[in] BSTR Name,
[in] BSTR Value,
[out, retval] VARIANT_BOOL* Supported);
[id(0x0000000f), hidden, helpstring("GetPropertyString")]
HRESULT GetPropertyString(
[in] BSTR Name,
[in, out] BSTR* Value,
[out, retval] VARIANT_BOOL* Supported);
[id(0x00000010), helpstring("IsUISupported")]
HRESULT IsUISupported(
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData,
[out, retval] VARIANT_BOOL* Supported);
[id(0x00000011), helpstring("DisplayUI")]
HRESULT DisplayUI(
[in] long hWndParent,
[in] BSTR Title,
[in] BSTR TypeOfUI,
[in, optional, defaultvalue("")] VARIANT* ExtraData);
[id(0x00000012), helpstring("GetRecognizers")]
HRESULT GetRecognizers(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** ObjectTokens);
[id(0x00000013), helpstring("GetAudioInputs")]
HRESULT GetAudioInputs(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** ObjectTokens);
[id(0x00000014), helpstring("GetProfiles")]
HRESULT GetProfiles(
[in, optional, defaultvalue("")] BSTR RequiredAttributes,
[in, optional, defaultvalue("")] BSTR OptionalAttributes,
[out, retval] ISpeechObjectTokens** ObjectTokens);
};
typedef enum {
SRSInactive = 0,
SRSActive = 1,
SRSActiveAlways = 2,
SRSInactiveWithPurge = 3
} SpeechRecognizerState;
[
odl,
uuid(BFF9E781-53EC-484E-BB8A-0E1B5551E35C),
helpstring("ISpeechRecognizerStatus Interface"),
dual,
oleautomation
]
interface ISpeechRecognizerStatus : IDispatch {
[id(0x00000001), propget, helpstring("AudioStatus")]
HRESULT AudioStatus([out, retval] ISpeechAudioStatus** AudioStatus);
[id(0x00000002), propget, helpstring("CurrentStreamPosition")]
HRESULT CurrentStreamPosition([out, retval] VARIANT* pCurrentStreamPos);
[id(0x00000003), propget, helpstring("CurrentStreamNumber")]
HRESULT CurrentStreamNumber([out, retval] long* StreamNumber);
[id(0x00000004), propget, helpstring("NumberOfActiveRules")]
HRESULT NumberOfActiveRules([out, retval] long* NumberOfActiveRules);
[id(0x00000005), propget, helpstring("ClsidEngine")]
HRESULT ClsidEngine([out, retval] BSTR* ClsidEngine);
[id(0x00000006), propget, helpstring("SupportedLanguages")]
HRESULT SupportedLanguages([out, retval] VARIANT* SupportedLanguages);
};
[
odl,
uuid(580AA49D-7E1E-4809-B8E2-57DA806104B8),
helpstring("ISpeechRecoContext Interface"),
dual,
oleautomation
]
interface ISpeechRecoContext : IDispatch {
[id(0x00000001), propget, helpstring("Recognizer")]
HRESULT Recognizer([out, retval] ISpeechRecognizer** Recognizer);
[id(0x00000002), propget, helpstring("AudioInInterferenceStatus")]
HRESULT AudioInputInterferenceStatus([out, retval] SpeechInterference* Interference);
[id(0x00000003), propget, helpstring("RequestedUIType")]
HRESULT RequestedUIType([out, retval] BSTR* UIType);
[id(0x00000004), propputref, helpstring("Voice")]
HRESULT Voice([in] ISpeechVoice* Voice);
[id(0x00000004), propget, helpstring("Voice")]
HRESULT Voice([out, retval] ISpeechVoice** Voice);
[id(0x00000005), propput, hidden, helpstring("AllowVoiceFormatMatchingOnNextSet")]
HRESULT AllowVoiceFormatMatchingOnNextSet([in] VARIANT_BOOL pAllow);
[id(0x00000005), propget, hidden, helpstring("AllowVoiceFormatMatchingOnNextSet")]
HRESULT AllowVoiceFormatMatchingOnNextSet([out, retval] VARIANT_BOOL* pAllow);
[id(0x00000006), propput, helpstring("VoicePurgeEvent")]
HRESULT VoicePurgeEvent([in] SpeechRecoEvents EventInterest);
[id(0x00000006), propget, helpstring("VoicePurgeEvent")]
HRESULT VoicePurgeEvent([out, retval] SpeechRecoEvents* EventInterest);
[id(0x00000007), propput, helpstring("EventInterests")]
HRESULT EventInterests([in] SpeechRecoEvents EventInterest);
[id(0x00000007), propget, helpstring("EventInterests")]
HRESULT EventInterests([out, retval] SpeechRecoEvents* EventInterest);
[id(0x00000008), propput, helpstring("CmdMaxAlternates")]
HRESULT CmdMaxAlternates([in] long MaxAlternates);
[id(0x00000008), propget, helpstring("CmdMaxAlternates")]
HRESULT CmdMaxAlternates([out, retval] long* MaxAlternates);
[id(0x00000009), propput, helpstring("State")]
HRESULT State([in] SpeechRecoContextState State);
[id(0x00000009), propget, helpstring("State")]
HRESULT State([out, retval] SpeechRecoContextState* State);
[id(0x0000000a), propput, helpstring("RetainedAudio")]
HRESULT RetainedAudio([in] SpeechRetainedAudioOptions Option);
[id(0x0000000a), propget, helpstring("RetainedAudio")]
HRESULT RetainedAudio([out, retval] SpeechRetainedAudioOptions* Option);
[id(0x0000000b), propputref, helpstring("RetainedAudioFormat")]
HRESULT RetainedAudioFormat([in] ISpeechAudioFormat* Format);
[id(0x0000000b), propget, helpstring("RetainedAudioFormat")]
HRESULT RetainedAudioFormat([out, retval] ISpeechAudioFormat** Format);
[id(0x0000000c), helpstring("Pause")]
HRESULT Pause();
[id(0x0000000d), helpstring("Resume")]
HRESULT Resume();
[id(0x0000000e), helpstring("CreateGrammar")]
HRESULT CreateGrammar(
[in, optional, defaultvalue(0)] VARIANT GrammarId,
[out, retval] ISpeechRecoGrammar** Grammar);
[id(0x0000000f), helpstring("CreateResultFromMemory")]
HRESULT CreateResultFromMemory(
[in] VARIANT* ResultBlock,
[out, retval] ISpeechRecoResult** Result);
[id(0x00000010), helpstring("Bookmark")]
HRESULT Bookmark(
[in] SpeechBookmarkOptions Options,
[in] VARIANT StreamPos,
[in] VARIANT BookmarkId);
[id(0x00000011), helpstring("SetAdaptationData")]
HRESULT SetAdaptationData([in] BSTR AdaptationString);
};
typedef enum {
SINone = 0,
SINoise = 1,
SINoSignal = 2,
SITooLoud = 3,
SITooQuiet = 4,
SITooFast = 5,
SITooSlow = 6
} SpeechInterference;
typedef enum {
SREStreamEnd = 1,
SRESoundStart = 2,
SRESoundEnd = 4,
SREPhraseStart = 8,
SRERecognition = 16,
SREHypothesis = 32,
SREBookmark = 64,
SREPropertyNumChange = 128,
SREPropertyStringChange = 256,
SREFalseRecognition = 512,
SREInterference = 1024,
SRERequestUI = 2048,
SREStateChange = 4096,
SREAdaptation = 8192,
SREStreamStart = 16384,
SRERecoOtherContext = 32768,
SREAudioLevel = 0x00010000,
SREPrivate = 0x00040000,
SREAllEvents = 0x0005ffff
} SpeechRecoEvents;
typedef enum {
SRCS_Disabled = 0,
SRCS_Enabled = 1
} SpeechRecoContextState;
typedef enum {
SRAONone = 0,
SRAORetainAudio = 1
} SpeechRetainedAudioOptions;
[
odl,
uuid(B6D6F79F-2158-4E50-B5BC-9A9CCD852A09),
helpstring("ISpeechRecoGrammar Interface"),
dual,
oleautomation
]
interface ISpeechRecoGrammar : IDispatch {
[id(0x00000001), propget, helpstring("Id")]
HRESULT Id([out, retval] VARIANT* Id);
[id(0x00000002), propget, helpstring("RecoContext")]
HRESULT RecoContext([out, retval] ISpeechRecoContext** RecoContext);
[id(0x00000003), propput, helpstring("State")]
HRESULT State([in] SpeechGrammarState State);
[id(0x00000003), propget, helpstring("State")]
HRESULT State([out, retval] SpeechGrammarState* State);
[id(0x00000004), propget, helpstring("Rules")]
HRESULT Rules([out, retval] ISpeechGrammarRules** Rules);
[id(0x00000005), helpstring("Reset")]
HRESULT Reset([in, optional, defaultvalue(0)] long NewLanguage);
[id(0x00000007), helpstring("CmdLoadFromFile")]
HRESULT CmdLoadFromFile(
[in] BSTR FileName,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x00000008), helpstring("CmdLoadFromObject")]
HRESULT CmdLoadFromObject(
[in] BSTR ClassId,
[in] BSTR GrammarName,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x00000009), helpstring("CmdLoadFromResource")]
HRESULT CmdLoadFromResource(
[in] long hModule,
[in] VARIANT ResourceName,
[in] VARIANT ResourceType,
[in] long LanguageId,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x0000000a), helpstring("CmdLoadFromMemory")]
HRESULT CmdLoadFromMemory(
[in] VARIANT GrammarData,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x0000000b), helpstring("CmdLoadFromProprietaryGrammar")]
HRESULT CmdLoadFromProprietaryGrammar(
[in] BSTR ProprietaryGuid,
[in] BSTR ProprietaryString,
[in] VARIANT ProprietaryData,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x0000000c), helpstring("CmdSetRuleState")]
HRESULT CmdSetRuleState(
[in] BSTR Name,
[in] SpeechRuleState State);
[id(0x0000000d), helpstring("CmdSetRuleIdState")]
HRESULT CmdSetRuleIdState(
[in] long RuleId,
[in] SpeechRuleState State);
[id(0x0000000e), helpstring("DictationLoad")]
HRESULT DictationLoad(
[in, optional, defaultvalue("")] BSTR TopicName,
[in, optional, defaultvalue(0)] SpeechLoadOption LoadOption);
[id(0x0000000f), helpstring("DictationUnload")]
HRESULT DictationUnload();
[id(0x00000010), helpstring("DictationSetState")]
HRESULT DictationSetState([in] SpeechRuleState State);
[id(0x00000011), helpstring("SetWordSequenceData")]
HRESULT SetWordSequenceData(
[in] BSTR Text,
[in] long TextLength,
[in] ISpeechTextSelectionInformation* Info);
[id(0x00000012), helpstring("SetTextSelection")]
HRESULT SetTextSelection([in] ISpeechTextSelectionInformation* Info);
[id(0x00000013), helpstring("IsPronounceable")]
HRESULT IsPronounceable(
[in] BSTR Word,
[out, retval] SpeechWordPronounceable* WordPronounceable);
};
typedef enum {
SGSEnabled = 1,
SGSDisabled = 0,
SGSExclusive = 3
} SpeechGrammarState;
[
odl,
uuid(6FFA3B44-FC2D-40D1-8AFC-32911C7F1AD1),
helpstring("ISpeechGrammarRules Interface"),
dual,
oleautomation
]
interface ISpeechGrammarRules : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(0x00000006), helpstring("FindRule")]
HRESULT FindRule(
[in] VARIANT RuleNameOrId,
[out, retval] ISpeechGrammarRule** Rule);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechGrammarRule** Rule);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the alternates")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
[id(0x00000002), propget, helpstring("Dynamic")]
HRESULT Dynamic([out, retval] VARIANT_BOOL* Dynamic);
[id(0x00000003), helpstring("Add")]
HRESULT Add(
[in] BSTR RuleName,
[in] SpeechRuleAttributes Attributes,
[in, optional, defaultvalue(0)] long RuleId,
[out, retval] ISpeechGrammarRule** Rule);
[id(0x00000004), helpstring("Commit")]
HRESULT Commit();
[id(0x00000005), helpstring("CommitAndSave")]
HRESULT CommitAndSave(
[out] BSTR* ErrorText,
[out, retval] VARIANT* SaveStream);
};
[
odl,
uuid(AFE719CF-5DD1-44F2-999C-7A399F1CFCCC),
helpstring("ISpeechGrammarRule Interface"),
dual,
oleautomation
]
interface ISpeechGrammarRule : IDispatch {
[id(0x00000001), propget, helpstring("RuleAttributes")]
HRESULT Attributes([out, retval] SpeechRuleAttributes* Attributes);
[id(0x00000002), propget, helpstring("InitialState")]
HRESULT InitialState([out, retval] ISpeechGrammarRuleState** State);
[id(0x00000003), propget, helpstring("Name")]
HRESULT Name([out, retval] BSTR* Name);
[id(0x00000004), propget, helpstring("Id")]
HRESULT Id([out, retval] long* Id);
[id(0x00000005), helpstring("Clear")]
HRESULT Clear();
[id(0x00000006), helpstring("AddResource")]
HRESULT AddResource(
[in] BSTR ResourceName,
[in] BSTR ResourceValue);
[id(0x00000007), helpstring("AddState")]
HRESULT AddState([out, retval] ISpeechGrammarRuleState** State);
};
typedef enum {
SRATopLevel = 1,
SRADefaultToActive = 2,
SRAExport = 4,
SRAImport = 8,
SRAInterpreter = 16,
SRADynamic = 32,
SRARoot = 64
} SpeechRuleAttributes;
[
odl,
uuid(D4286F2C-EE67-45AE-B928-28D695362EDA),
helpstring("ISpeechGrammarRuleState Interface"),
dual,
oleautomation
]
interface ISpeechGrammarRuleState : IDispatch {
[id(0x00000001), propget, helpstring("Rule")]
HRESULT Rule([out, retval] ISpeechGrammarRule** Rule);
[id(0x00000002), propget, helpstring("Transitions")]
HRESULT Transitions([out, retval] ISpeechGrammarRuleStateTransitions** Transitions);
[id(0x00000003), helpstring("AddWordTransition")]
HRESULT AddWordTransition(
[in] ISpeechGrammarRuleState* DestState,
[in] BSTR Words,
[in, optional, defaultvalue(" ")] BSTR Separators,
[in, optional, defaultvalue(1)] SpeechGrammarWordType Type,
[in, optional, defaultvalue("")] BSTR PropertyName,
[in, optional, defaultvalue(0)] long PropertyId,
[in, optional, defaultvalue("")] VARIANT* PropertyValue,
[in, optional, defaultvalue(1)] single Weight);
[id(0x00000004), helpstring("AddRuleTransition")]
HRESULT AddRuleTransition(
[in] ISpeechGrammarRuleState* DestinationState,
[in] ISpeechGrammarRule* Rule,
[in, optional, defaultvalue("")] BSTR PropertyName,
[in, optional, defaultvalue(0)] long PropertyId,
[in, optional, defaultvalue("")] VARIANT* PropertyValue,
[in, optional, defaultvalue(1)] single Weight);
[id(0x00000005), helpstring("AddSpecialTransition")]
HRESULT AddSpecialTransition(
[in] ISpeechGrammarRuleState* DestinationState,
[in] SpeechSpecialTransitionType Type,
[in, optional, defaultvalue("")] BSTR PropertyName,
[in, optional, defaultvalue(0)] long PropertyId,
[in, optional, defaultvalue("")] VARIANT* PropertyValue,
[in, optional, defaultvalue(1)] single Weight);
};
[
odl,
uuid(EABCE657-75BC-44A2-AA7F-C56476742963),
helpstring("ISpeechGrammarRuleStateTransitions Interface"),
dual,
oleautomation
]
interface ISpeechGrammarRuleStateTransitions : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechGrammarRuleStateTransition** Transition);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the transitions")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(CAFD1DB1-41D1-4A06-9863-E2E81DA17A9A),
helpstring("ISpeechGrammarRuleStateTransition Interface"),
dual,
oleautomation
]
interface ISpeechGrammarRuleStateTransition : IDispatch {
[id(0x00000001), propget, helpstring("Type")]
HRESULT Type([out, retval] SpeechGrammarRuleStateTransitionType* Type);
[id(0x00000002), propget, helpstring("Text")]
HRESULT Text([out, retval] BSTR* Text);
[id(0x00000003), propget, helpstring("Rule")]
HRESULT Rule([out, retval] ISpeechGrammarRule** Rule);
[id(0x00000004), propget, helpstring("Weight")]
HRESULT Weight([out, retval] VARIANT* Weight);
[id(0x00000005), propget, helpstring("PropertyName")]
HRESULT PropertyName([out, retval] BSTR* PropertyName);
[id(0x00000006), propget, helpstring("PropertyId")]
HRESULT PropertyId([out, retval] long* PropertyId);
[id(0x00000007), propget, helpstring("PropertyValue")]
HRESULT PropertyValue([out, retval] VARIANT* PropertyValue);
[id(0x00000008), propget, helpstring("NextState")]
HRESULT NextState([out, retval] ISpeechGrammarRuleState** NextState);
};
typedef enum {
SGRSTTEpsilon = 0,
SGRSTTWord = 1,
SGRSTTRule = 2,
SGRSTTDictation = 3,
SGRSTTWildcard = 4,
SGRSTTTextBuffer = 5
} SpeechGrammarRuleStateTransitionType;
typedef enum {
SGDisplay = 0,
SGLexical = 1,
SGPronounciation = 2,
SGLexicalNoSpecialChars = 3
} SpeechGrammarWordType;
typedef enum {
SSTTWildcard = 1,
SSTTDictation = 2,
SSTTTextBuffer = 3
} SpeechSpecialTransitionType;
typedef enum {
SLOStatic = 0,
SLODynamic = 1
} SpeechLoadOption;
typedef enum {
SGDSInactive = 0,
SGDSActive = 1,
SGDSActiveWithAutoPause = 3,
SGDSActiveUserDelimited = 4
} SpeechRuleState;
[
odl,
uuid(3B9C7E7A-6EEE-4DED-9092-11657279ADBE),
helpstring("ISpeechTextSelectionInformation Interface"),
dual,
oleautomation
]
interface ISpeechTextSelectionInformation : IDispatch {
[id(0x00000001), propput, helpstring("ActiveOffset")]
HRESULT ActiveOffset([in] long ActiveOffset);
[id(0x00000001), propget, helpstring("ActiveOffset")]
HRESULT ActiveOffset([out, retval] long* ActiveOffset);
[id(0x00000002), propput, helpstring("ActiveLength")]
HRESULT ActiveLength([in] long ActiveLength);
[id(0x00000002), propget, helpstring("ActiveLength")]
HRESULT ActiveLength([out, retval] long* ActiveLength);
[id(0x00000003), propput, helpstring("SelectionOffset")]
HRESULT SelectionOffset([in] long SelectionOffset);
[id(0x00000003), propget, helpstring("SelectionOffset")]
HRESULT SelectionOffset([out, retval] long* SelectionOffset);
[id(0x00000004), propput, helpstring("SelectionLength")]
HRESULT SelectionLength([in] long SelectionLength);
[id(0x00000004), propget, helpstring("SelectionLength")]
HRESULT SelectionLength([out, retval] long* SelectionLength);
};
typedef enum {
SWPUnknownWordUnpronounceable = 0,
SWPUnknownWordPronounceable = 1,
SWPKnownWordPronounceable = 2
} SpeechWordPronounceable;
[
odl,
uuid(ED2879CF-CED9-4EE6-A534-DE0191D5468D),
helpstring("ISpeechRecoResult Interface"),
dual,
oleautomation
]
interface ISpeechRecoResult : IDispatch {
[id(0x00000001), propget, helpstring("RecoContext")]
HRESULT RecoContext([out, retval] ISpeechRecoContext** RecoContext);
[id(0x00000002), propget, helpstring("Times")]
HRESULT Times([out, retval] ISpeechRecoResultTimes** Times);
[id(0x00000003), propputref, helpstring("AudioFormat")]
HRESULT AudioFormat([in] ISpeechAudioFormat* Format);
[id(0x00000003), propget, helpstring("AudioFormat")]
HRESULT AudioFormat([out, retval] ISpeechAudioFormat** Format);
[id(0x00000004), propget, helpstring("PhraseInfo")]
HRESULT PhraseInfo([out, retval] ISpeechPhraseInfo** PhraseInfo);
[id(0x00000005), helpstring("Alternates")]
HRESULT Alternates(
[in] long RequestCount,
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[out, retval] ISpeechPhraseAlternates** Alternates);
[id(0x00000006), helpstring("Audio")]
HRESULT Audio(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[out, retval] ISpeechMemoryStream** Stream);
[id(0x00000007), helpstring("SpeakAudio")]
HRESULT SpeakAudio(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[in, optional, defaultvalue(0)] SpeechVoiceSpeakFlags Flags,
[out, retval] long* StreamNumber);
[id(0x00000008), helpstring("SaveToMemory")]
HRESULT SaveToMemory([out, retval] VARIANT* ResultBlock);
[id(0x00000009), helpstring("DiscardResultInfo")]
HRESULT DiscardResultInfo([in] SpeechDiscardType ValueTypes);
};
[
odl,
uuid(62B3B8FB-F6E7-41BE-BDCB-056B1C29EFC0),
helpstring("ISpeechRecoResultTimes Interface"),
dual,
oleautomation
]
interface ISpeechRecoResultTimes : IDispatch {
[id(0x00000001), propget, helpstring("StreamTime")]
HRESULT StreamTime([out, retval] VARIANT* Time);
[id(0x00000002), propget, helpstring("Length")]
HRESULT Length([out, retval] VARIANT* Length);
[id(0x00000003), propget, helpstring("TickCount")]
HRESULT TickCount([out, retval] long* TickCount);
[id(0x00000004), propget, helpstring("Start")]
HRESULT OffsetFromStart([out, retval] VARIANT* OffsetFromStart);
};
[
odl,
uuid(961559CF-4E67-4662-8BF0-D93F1FCD61B3),
helpstring("ISpeechPhraseInfo Interface"),
dual,
oleautomation
]
interface ISpeechPhraseInfo : IDispatch {
[id(0x00000001), propget, helpstring("LanguageId")]
HRESULT LanguageId([out, retval] long* LanguageId);
[id(0x00000002), propget, helpstring("GrammarId")]
HRESULT GrammarId([out, retval] VARIANT* GrammarId);
[id(0x00000003), propget, helpstring("StartTime")]
HRESULT StartTime([out, retval] VARIANT* StartTime);
[id(0x00000004), propget, helpstring("AudioStreamPosition")]
HRESULT AudioStreamPosition([out, retval] VARIANT* AudioStreamPosition);
[id(0x00000005), propget, helpstring("AudioSizeBytes")]
HRESULT AudioSizeBytes([out, retval] long* pAudioSizeBytes);
[id(0x00000006), propget, helpstring("RetainedSizeBytes")]
HRESULT RetainedSizeBytes([out, retval] long* RetainedSizeBytes);
[id(0x00000007), propget, helpstring("AudioSizeTime")]
HRESULT AudioSizeTime([out, retval] long* AudioSizeTime);
[id(0x00000008), propget, helpstring("Rule")]
HRESULT Rule([out, retval] ISpeechPhraseRule** Rule);
[id(0x00000009), propget, helpstring("Properties")]
HRESULT Properties([out, retval] ISpeechPhraseProperties** Properties);
[id(0x0000000a), propget, helpstring("Elements")]
HRESULT Elements([out, retval] ISpeechPhraseElements** Elements);
[id(0x0000000b), propget, helpstring("Replacements")]
HRESULT Replacements([out, retval] ISpeechPhraseReplacements** Replacements);
[id(0x0000000c), propget, helpstring("EngineId")]
HRESULT EngineId([out, retval] BSTR* EngineIdGuid);
[id(0x0000000d), propget, helpstring("EnginePrivateData")]
HRESULT EnginePrivateData([out, retval] VARIANT* PrivateData);
[id(0x0000000e), helpstring("SaveToMemory")]
HRESULT SaveToMemory([out, retval] VARIANT* PhraseBlock);
[id(0x0000000f), helpstring("GetText")]
HRESULT GetText(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[in, optional, defaultvalue(-1)] VARIANT_BOOL UseReplacements,
[out, retval] BSTR* Text);
[id(0x00000010), helpstring("DisplayAttributes")]
HRESULT GetDisplayAttributes(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[in, optional, defaultvalue(-1)] VARIANT_BOOL UseReplacements,
[out, retval] SpeechDisplayAttributes* DisplayAttributes);
};
[
odl,
uuid(A7BFE112-A4A0-48D9-B602-C313843F6964),
helpstring("ISpeechPhraseRule Interface"),
dual,
oleautomation
]
interface ISpeechPhraseRule : IDispatch {
[id(0x00000001), propget, helpstring("Name")]
HRESULT Name([out, retval] BSTR* Name);
[id(0x00000002), propget, helpstring("Id")]
HRESULT Id([out, retval] long* Id);
[id(0x00000003), propget, helpstring("FirstElement")]
HRESULT FirstElement([out, retval] long* FirstElement);
[id(0x00000004), propget, helpstring("NumElements")]
HRESULT NumberOfElements([out, retval] long* NumberOfElements);
[id(0x00000005), propget, helpstring("Parent")]
HRESULT Parent([out, retval] ISpeechPhraseRule** Parent);
[id(0x00000006), propget, helpstring("Children")]
HRESULT Children([out, retval] ISpeechPhraseRules** Children);
[id(0x00000007), propget, helpstring("Confidence")]
HRESULT Confidence([out, retval] SpeechEngineConfidence* ActualConfidence);
[id(0x00000008), propget, helpstring("EngineConfidence")]
HRESULT EngineConfidence([out, retval] single* EngineConfidence);
};
[
odl,
uuid(9047D593-01DD-4B72-81A3-E4A0CA69F407),
helpstring("ISpeechPhraseRules Interface"),
dual,
oleautomation
]
interface ISpeechPhraseRules : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechPhraseRule** Rule);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the Rules")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
typedef enum {
SECLowConfidence = 0xffffffff,
SECNormalConfidence = 0,
SECHighConfidence = 1
} SpeechEngineConfidence;
[
odl,
uuid(08166B47-102E-4B23-A599-BDB98DBFD1F4),
helpstring("ISpeechPhraseProperties Interface"),
dual,
oleautomation
]
interface ISpeechPhraseProperties : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechPhraseProperty** Property);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the alternates")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(CE563D48-961E-4732-A2E1-378A42B430BE),
helpstring("ISpeechPhraseProperty Interface"),
dual,
oleautomation
]
interface ISpeechPhraseProperty : IDispatch {
[id(0x00000001), propget, helpstring("Name")]
HRESULT Name([out, retval] BSTR* Name);
[id(0x00000002), propget, helpstring("Id")]
HRESULT Id([out, retval] long* Id);
[id(0x00000003), propget, helpstring("Value")]
HRESULT Value([out, retval] VARIANT* Value);
[id(0x00000004), propget, helpstring("FirstElement")]
HRESULT FirstElement([out, retval] long* FirstElement);
[id(0x00000005), propget, helpstring("NumberOfElements")]
HRESULT NumberOfElements([out, retval] long* NumberOfElements);
[id(0x00000006), propget, helpstring("EngineConfidence")]
HRESULT EngineConfidence([out, retval] single* Confidence);
[id(0x00000007), propget, helpstring("Confidence")]
HRESULT Confidence([out, retval] SpeechEngineConfidence* Confidence);
[id(0x00000008), propget, helpstring("Parent")]
HRESULT Parent([out, retval] ISpeechPhraseProperty** ParentProperty);
[id(0x00000009), propget, helpstring("Children")]
HRESULT Children([out, retval] ISpeechPhraseProperties** Children);
};
[
odl,
uuid(0626B328-3478-467D-A0B3-D0853B93DDA3),
helpstring("ISpeechPhraseElements Interface"),
dual,
oleautomation
]
interface ISpeechPhraseElements : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechPhraseElement** Element);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the tokens")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(E6176F96-E373-4801-B223-3B62C068C0B4),
helpstring("ISpeechPhraseElement Interface"),
dual,
oleautomation
]
interface ISpeechPhraseElement : IDispatch {
[id(0x00000001), propget, helpstring("AudioTimeOffset")]
HRESULT AudioTimeOffset([out, retval] long* AudioTimeOffset);
[id(0x00000002), propget, helpstring("AudioSizeTime")]
HRESULT AudioSizeTime([out, retval] long* AudioSizeTime);
[id(0x00000003), propget, helpstring("AudioStreamOffset")]
HRESULT AudioStreamOffset([out, retval] long* AudioStreamOffset);
[id(0x00000004), propget, helpstring("AudioSizeBytes")]
HRESULT AudioSizeBytes([out, retval] long* AudioSizeBytes);
[id(0x00000005), propget, helpstring("RetainedStreamOffset")]
HRESULT RetainedStreamOffset([out, retval] long* RetainedStreamOffset);
[id(0x00000006), propget, helpstring("RetainedSizeBytes")]
HRESULT RetainedSizeBytes([out, retval] long* RetainedSizeBytes);
[id(0x00000007), propget, helpstring("DisplayText")]
HRESULT DisplayText([out, retval] BSTR* DisplayText);
[id(0x00000008), propget, helpstring("LexicalForm")]
HRESULT LexicalForm([out, retval] BSTR* LexicalForm);
[id(0x00000009), propget, helpstring("Pronunciation")]
HRESULT Pronunciation([out, retval] VARIANT* Pronunciation);
[id(0x0000000a), propget, helpstring("DisplayAttributes")]
HRESULT DisplayAttributes([out, retval] SpeechDisplayAttributes* DisplayAttributes);
[id(0x0000000b), propget, helpstring("RequiredConfidence")]
HRESULT RequiredConfidence([out, retval] SpeechEngineConfidence* RequiredConfidence);
[id(0x0000000c), propget, helpstring("ActualConfidence")]
HRESULT ActualConfidence([out, retval] SpeechEngineConfidence* ActualConfidence);
[id(0x0000000d), propget, helpstring("EngineConfidence")]
HRESULT EngineConfidence([out, retval] single* EngineConfidence);
};
typedef enum {
SDA_No_Trailing_Space = 0,
SDA_One_Trailing_Space = 2,
SDA_Two_Trailing_Spaces = 4,
SDA_Consume_Leading_Spaces = 8
} SpeechDisplayAttributes;
[
odl,
uuid(38BC662F-2257-4525-959E-2069D2596C05),
helpstring("ISpeechPhraseReplacements Interface"),
dual,
oleautomation
]
interface ISpeechPhraseReplacements : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechPhraseReplacement** Reps);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the tokens")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(2890A410-53A7-4FB5-94EC-06D4998E3D02),
helpstring("ISpeechPhraseReplacement Interface"),
dual,
oleautomation
]
interface ISpeechPhraseReplacement : IDispatch {
[id(0x00000001), propget, helpstring("DisplayAttributes")]
HRESULT DisplayAttributes([out, retval] SpeechDisplayAttributes* DisplayAttributes);
[id(0x00000002), propget, helpstring("Text")]
HRESULT Text([out, retval] BSTR* Text);
[id(0x00000003), propget, helpstring("FirstElement")]
HRESULT FirstElement([out, retval] long* FirstElement);
[id(0x00000004), propget, helpstring("NumElements")]
HRESULT NumberOfElements([out, retval] long* NumberOfElements);
};
[
odl,
uuid(B238B6D5-F276-4C3D-A6C1-2974801C3CC2),
helpstring("ISpeechPhraseAlternates Interface"),
dual,
oleautomation
]
interface ISpeechPhraseAlternates : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechPhraseAlternate** PhraseAlternate);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the alternates")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(27864A2A-2B9F-4CB8-92D3-0D2722FD1E73),
helpstring("ISpeechPhraseAlternate Interface"),
dual,
oleautomation
]
interface ISpeechPhraseAlternate : IDispatch {
[id(0x00000001), propget, helpstring("RecoResult")]
HRESULT RecoResult([out, retval] ISpeechRecoResult** RecoResult);
[id(0x00000002), propget, helpstring("StartElementInResult")]
HRESULT StartElementInResult([out, retval] long* StartElement);
[id(0x00000003), propget, helpstring("NumberOfElementsInResult")]
HRESULT NumberOfElementsInResult([out, retval] long* NumberOfElements);
[id(0x00000004), propget, helpstring("Phrase")]
HRESULT PhraseInfo([out, retval] ISpeechPhraseInfo** PhraseInfo);
[id(0x00000005), helpstring("Commit")]
HRESULT Commit();
};
typedef enum {
SDTProperty = 1,
SDTReplacement = 2,
SDTRule = 4,
SDTDisplayText = 8,
SDTLexicalForm = 16,
SDTPronunciation = 32,
SDTAudio = 64,
SDTAlternates = 128,
SDTAll = 255
} SpeechDiscardType;
typedef enum {
SBONone = 0,
SBOPause = 1
} SpeechBookmarkOptions;
typedef enum {
SFTInput = 0,
SFTSREngine = 1
} SpeechFormatType;
[
uuid(7B8FCB42-0E9D-4F00-A048-7B04D6179D3D)
]
dispinterface _ISpeechRecoContextEvents {
properties:
methods:
[id(0x00000001), helpstring("StartStream")]
void StartStream(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000002), helpstring("EndStream")]
void EndStream(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] VARIANT_BOOL StreamReleased);
[id(0x00000003), helpstring("Bookmark")]
void Bookmark(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] VARIANT BookmarkId,
[in] SpeechBookmarkOptions Options);
[id(0x00000004), helpstring("SoundStart")]
void SoundStart(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000005), helpstring("SoundEnd")]
void SoundEnd(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000006), helpstring("PhraseStart")]
void PhraseStart(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000007), helpstring("Recognition")]
void Recognition(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] SpeechRecognitionType RecognitionType,
[in] ISpeechRecoResult* Result);
[id(0x00000008), helpstring("Hypothesis")]
void Hypothesis(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] ISpeechRecoResult* Result);
[id(0x00000009), helpstring("PropertyNumberChange")]
void PropertyNumberChange(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] BSTR PropertyName,
[in] long NewNumberValue);
[id(0x0000000a), helpstring("PropertyStringChange")]
void PropertyStringChange(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] BSTR PropertyName,
[in] BSTR NewStringValue);
[id(0x0000000b), helpstring("FalseRecognition")]
void FalseRecognition(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] ISpeechRecoResult* Result);
[id(0x0000000c), helpstring("Interference")]
void Interference(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] SpeechInterference Interference);
[id(0x0000000d), helpstring("RequestUI")]
void RequestUI(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] BSTR UIType);
[id(0x0000000e), helpstring("RecognizerStateChange")]
void RecognizerStateChange(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] SpeechRecognizerState NewState);
[id(0x0000000f), helpstring("Adaptation")]
void Adaptation(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000010), helpstring("RecognitionForOtherContext")]
void RecognitionForOtherContext(
[in] long StreamNumber,
[in] VARIANT StreamPosition);
[id(0x00000011), helpstring("AudioLevel")]
void AudioLevel(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] long AudioLevel);
[id(0x00000012), helpstring("EnginePrivate")]
void EnginePrivate(
[in] long StreamNumber,
[in] VARIANT StreamPosition,
[in] VARIANT EngineData);
};
typedef enum {
SRTStandard = 0,
SRTAutopause = 1,
SRTEmulated = 2,
SRTSMLTimeout = 4,
SRTExtendableParse = 8,
SRTReSent = 16
} SpeechRecognitionType;
[
odl,
uuid(8E0A246D-D3C8-45DE-8657-04290C458C3C),
helpstring("ISpeechRecoResult2 Interface"),
dual,
oleautomation
]
interface ISpeechRecoResult2 : ISpeechRecoResult {
[id(0x0000000c), helpstring("DiscardResultInfo")]
HRESULT SetTextFeedback(
[in] BSTR Feedback,
[in] VARIANT_BOOL WasSuccessful);
};
[
odl,
uuid(3DA7627A-C7AE-4B23-8708-638C50362C25),
helpstring("ISpeechLexicon Interface"),
dual,
oleautomation
]
interface ISpeechLexicon : IDispatch {
[id(0x00000001), propget, hidden, helpstring("GenerationId")]
HRESULT GenerationId([out, retval] long* GenerationId);
[id(0x00000002), helpstring("GetWords")]
HRESULT GetWords(
[in, optional, defaultvalue(3)] SpeechLexiconType Flags,
[out, optional, defaultvalue(0)] long* GenerationId,
[out, retval] ISpeechLexiconWords** Words);
[id(0x00000003), helpstring("AddPronunciation")]
HRESULT AddPronunciation(
[in] BSTR bstrWord,
[in] long LangId,
[in, optional, defaultvalue(0)] SpeechPartOfSpeech PartOfSpeech,
[in, optional, defaultvalue("")] BSTR bstrPronunciation);
[id(0x00000004), hidden, helpstring("AddPronunciationByPhoneIds")]
HRESULT AddPronunciationByPhoneIds(
[in] BSTR bstrWord,
[in] long LangId,
[in, optional, defaultvalue(0)] SpeechPartOfSpeech PartOfSpeech,
[in, optional, defaultvalue("")] VARIANT* PhoneIds);
[id(0x00000005), helpstring("RemovePronunciation")]
HRESULT RemovePronunciation(
[in] BSTR bstrWord,
[in] long LangId,
[in, optional, defaultvalue(0)] SpeechPartOfSpeech PartOfSpeech,
[in, optional, defaultvalue("")] BSTR bstrPronunciation);
[id(0x00000006), hidden, helpstring("RemovePronunciationByPhoneIds")]
HRESULT RemovePronunciationByPhoneIds(
[in] BSTR bstrWord,
[in] long LangId,
[in, optional, defaultvalue(0)] SpeechPartOfSpeech PartOfSpeech,
[in, optional, defaultvalue("")] VARIANT* PhoneIds);
[id(0x00000007), helpstring("GetPronunciations")]
HRESULT GetPronunciations(
[in] BSTR bstrWord,
[in, optional, defaultvalue(0)] long LangId,
[in, optional, defaultvalue(3)] SpeechLexiconType TypeFlags,
[out, retval] ISpeechLexiconPronunciations** ppPronunciations);
[id(0x00000008), hidden, helpstring("GetGenerationChange")]
HRESULT GetGenerationChange(
[in, out] long* GenerationId,
[out, retval] ISpeechLexiconWords** ppWords);
};
typedef enum {
SLTUser = 1,
SLTApp = 2
} SpeechLexiconType;
[
odl,
uuid(8D199862-415E-47D5-AC4F-FAA608B424E6),
helpstring("ISpeechLexiconWords Interface"),
dual,
oleautomation
]
interface ISpeechLexiconWords : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechLexiconWord** Word);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the tokens")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(4E5B933C-C9BE-48ED-8842-1EE51BB1D4FF),
helpstring("ISpeechLexiconWord Interface"),
dual,
oleautomation
]
interface ISpeechLexiconWord : IDispatch {
[id(0x00000001), propget]
HRESULT LangId([out, retval] long* LangId);
[id(0x00000002), propget]
HRESULT Type([out, retval] SpeechWordType* WordType);
[id(0x00000003), propget]
HRESULT Word([out, retval] BSTR* Word);
[id(0x00000004), propget]
HRESULT Pronunciations([out, retval] ISpeechLexiconPronunciations** Pronunciations);
};
typedef enum {
SWTAdded = 1,
SWTDeleted = 2
} SpeechWordType;
[
odl,
uuid(72829128-5682-4704-A0D4-3E2BB6F2EAD3),
helpstring("ISpeechLexiconPronunciations Interface"),
dual,
oleautomation
]
interface ISpeechLexiconPronunciations : IDispatch {
[id(0x00000001), propget, helpstring("Count")]
HRESULT Count([out, retval] long* Count);
[id(00000000), helpstring("Item")]
HRESULT Item(
[in] long Index,
[out, retval] ISpeechLexiconPronunciation** Pronunciation);
[id(0xfffffffc), propget, restricted, helpstring("Enumerates the tokens")]
HRESULT _NewEnum([out, retval] IUnknown** EnumVARIANT);
};
[
odl,
uuid(95252C5D-9E43-4F4A-9899-48EE73352F9F),
helpstring("ISpeechLexiconPronunciation Interface"),
dual,
oleautomation
]
interface ISpeechLexiconPronunciation : IDispatch {
[id(0x00000001), propget, helpstring("Type")]
HRESULT Type([out, retval] SpeechLexiconType* LexiconType);
[id(0x00000002), propget, helpstring("LangId")]
HRESULT LangId([out, retval] long* LangId);
[id(0x00000003), propget, helpstring("PartOfSpeech")]
HRESULT PartOfSpeech([out, retval] SpeechPartOfSpeech* PartOfSpeech);
[id(0x00000004), propget, helpstring("PhoneIds")]
HRESULT PhoneIds([out, retval] VARIANT* PhoneIds);
[id(0x00000005), propget, helpstring("Symbolic")]
HRESULT Symbolic([out, retval] BSTR* Symbolic);
};
typedef enum {
SPSNotOverriden = 0xffffffff,
SPSUnknown = 0,
SPSNoun = 4096,
SPSVerb = 8192,
SPSModifier = 12288,
SPSFunction = 16384,
SPSInterjection = 20480,
SPSLMA = 28672,
SPSSuppressWord = 61440
} SpeechPartOfSpeech;
// NOTE: This module has no entry points. There is no way to
// extract the dllname of a module with no entry points!
//
[
dllname("<no entry points>"),
uuid(E58442E4-0C80-402C-9559-867337A39765)
]
module SpeechStringConstants {
const BSTR SpeechRegistryUserRoot = "HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Speech";
const BSTR SpeechRegistryLocalMachineRoot = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech";
const BSTR SpeechCategoryAudioOut = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\AudioOutput";
const BSTR SpeechCategoryAudioIn = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\AudioInput";
const BSTR SpeechCategoryVoices = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\Voices";
const BSTR SpeechCategoryRecognizers = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\Recognizers";
const BSTR SpeechCategoryAppLexicons = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\AppLexicons";
const BSTR SpeechCategoryPhoneConverters = "HKEY_LOCAL_MACHINE\\SOFTWARE\\Microsoft\\Speech\\PhoneConverters";
const BSTR SpeechCategoryRecoProfiles = "HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Speech\\RecoProfiles";
const BSTR SpeechTokenIdUserLexicon = "HKEY_CURRENT_USER\\SOFTWARE\\Microsoft\\Speech\\CurrentUserLexicon";
const BSTR SpeechTokenValueCLSID = "CLSID";
const BSTR SpeechTokenKeyFiles = "Files";
const BSTR SpeechTokenKeyUI = "UI";
const BSTR SpeechTokenKeyAttributes = "Attributes";
const BSTR SpeechVoiceCategoryTTSRate = "DefaultTTSRate";
const BSTR SpeechPropertyResourceUsage = "ResourceUsage";
const BSTR SpeechPropertyHighConfidenceThreshold = "HighConfidenceThreshold";
const BSTR SpeechPropertyNormalConfidenceThreshold = "NormalConfidenceThreshold";
const BSTR SpeechPropertyLowConfidenceThreshold = "LowConfidenceThreshold";
const BSTR SpeechPropertyResponseSpeed = "ResponseSpeed";
const BSTR SpeechPropertyComplexResponseSpeed = "ComplexResponseSpeed";
const BSTR SpeechPropertyAdaptationOn = "AdaptationOn";
const BSTR SpeechDictationTopicSpelling = "Spelling";
const BSTR SpeechGrammarTagWildcard = "...";
const BSTR SpeechGrammarTagDictation = "*";
const BSTR SpeechGrammarTagUnlimitedDictation = "*+";
const BSTR SpeechEngineProperties = "EngineProperties";
const BSTR SpeechAddRemoveWord = "AddRemoveWord";
const BSTR SpeechUserTraining = "UserTraining";
const BSTR SpeechMicTraining = "MicTraining";
const BSTR SpeechRecoProfileProperties = "RecoProfileProperties";
const BSTR SpeechAudioProperties = "AudioProperties";
const BSTR SpeechAudioVolume = "AudioVolume";
const BSTR SpeechVoiceSkipTypeSentence = "Sentence";
const BSTR SpeechAudioFormatGUIDWave = "{C31ADBAE-527F-4ff5-A230-F62BB61FF70C}";
const BSTR SpeechAudioFormatGUIDText = "{7CEEF9F9-3D13-11d2-9EE7-00C04F797396}";
};
// NOTE: This module has no entry points. There is no way to
// extract the dllname of a module with no entry points!
//
[
dllname("<no entry points>"),
uuid(F3E092B2-6BDC-410F-BCB2-4C5ED4424180)
]
module SpeechConstants {
const single Speech_Default_Weight = 1;
const long Speech_Max_Word_Length = 128;
const long Speech_Max_Pron_Length = 384;
const long Speech_StreamPos_Asap = 0;
const long Speech_StreamPos_RealTime = 0xffffffff;
const long SpeechAllElements = 0xffffffff;
};
typedef enum {
DISPID_SDKSetBinaryValue = 1,
DISPID_SDKGetBinaryValue = 2,
DISPID_SDKSetStringValue = 3,
DISPID_SDKGetStringValue = 4,
DISPID_SDKSetLongValue = 5,
DISPID_SDKGetlongValue = 6,
DISPID_SDKOpenKey = 7,
DISPID_SDKCreateKey = 8,
DISPID_SDKDeleteKey = 9,
DISPID_SDKDeleteValue = 10,
DISPID_SDKEnumKeys = 11,
DISPID_SDKEnumValues = 12
} DISPID_SpeechDataKey;
typedef enum {
DISPID_SOTId = 1,
DISPID_SOTDataKey = 2,
DISPID_SOTCategory = 3,
DISPID_SOTGetDescription = 4,
DISPID_SOTSetId = 5,
DISPID_SOTGetAttribute = 6,
DISPID_SOTCreateInstance = 7,
DISPID_SOTRemove = 8,
DISPID_SOTGetStorageFileName = 9,
DISPID_SOTRemoveStorageFileName = 10,
DISPID_SOTIsUISupported = 11,
DISPID_SOTDisplayUI = 12,
DISPID_SOTMatchesAttributes = 13
} DISPID_SpeechObjectToken;
typedef enum {
DISPID_SOTsCount = 1,
DISPID_SOTsItem = 0,
DISPID_SOTs_NewEnum = 0xfffffffc
} DISPID_SpeechObjectTokens;
typedef enum {
DISPID_SOTCId = 1,
DISPID_SOTCDefault = 2,
DISPID_SOTCSetId = 3,
DISPID_SOTCGetDataKey = 4,
DISPID_SOTCEnumerateTokens = 5
} DISPID_SpeechObjectTokenCategory;
typedef enum {
DISPID_SAFType = 1,
DISPID_SAFGuid = 2,
DISPID_SAFGetWaveFormatEx = 3,
DISPID_SAFSetWaveFormatEx = 4
} DISPID_SpeechAudioFormat;
typedef enum {
DISPID_SBSFormat = 1,
DISPID_SBSRead = 2,
DISPID_SBSWrite = 3,
DISPID_SBSSeek = 4
} DISPID_SpeechBaseStream;
typedef enum {
DISPID_SAStatus = 200,
DISPID_SABufferInfo = 201,
DISPID_SADefaultFormat = 202,
DISPID_SAVolume = 203,
DISPID_SABufferNotifySize = 204,
DISPID_SAEventHandle = 205,
DISPID_SASetState = 206
} DISPID_SpeechAudio;
typedef enum {
DISPID_SMSADeviceId = 300,
DISPID_SMSALineId = 301,
DISPID_SMSAMMHandle = 302
} DISPID_SpeechMMSysAudio;
typedef enum {
DISPID_SFSOpen = 100,
DISPID_SFSClose = 101
} DISPID_SpeechFileStream;
typedef enum {
DISPID_SCSBaseStream = 100
} DISPID_SpeechCustomStream;
typedef enum {
DISPID_SMSSetData = 100,
DISPID_SMSGetData = 101
} DISPID_SpeechMemoryStream;
typedef enum {
DISPID_SASFreeBufferSpace = 1,
DISPID_SASNonBlockingIO = 2,
DISPID_SASState = 3,
DISPID_SASCurrentSeekPosition = 4,
DISPID_SASCurrentDevicePosition = 5
} DISPID_SpeechAudioStatus;
typedef enum {
DISPID_SABIMinNotification = 1,
DISPID_SABIBufferSize = 2,
DISPID_SABIEventBias = 3
} DISPID_SpeechAudioBufferInfo;
typedef enum {
DISPID_SWFEFormatTag = 1,
DISPID_SWFEChannels = 2,
DISPID_SWFESamplesPerSec = 3,
DISPID_SWFEAvgBytesPerSec = 4,
DISPID_SWFEBlockAlign = 5,
DISPID_SWFEBitsPerSample = 6,
DISPID_SWFEExtraData = 7
} DISPID_SpeechWaveFormatEx;
typedef enum {
DISPID_SVStatus = 1,
DISPID_SVVoice = 2,
DISPID_SVAudioOutput = 3,
DISPID_SVAudioOutputStream = 4,
DISPID_SVRate = 5,
DISPID_SVVolume = 6,
DISPID_SVAllowAudioOuputFormatChangesOnNextSet = 7,
DISPID_SVEventInterests = 8,
DISPID_SVPriority = 9,
DISPID_SVAlertBoundary = 10,
DISPID_SVSyncronousSpeakTimeout = 11,
DISPID_SVSpeak = 12,
DISPID_SVSpeakStream = 13,
DISPID_SVPause = 14,
DISPID_SVResume = 15,
DISPID_SVSkip = 16,
DISPID_SVGetVoices = 17,
DISPID_SVGetAudioOutputs = 18,
DISPID_SVWaitUntilDone = 19,
DISPID_SVSpeakCompleteEvent = 20,
DISPID_SVIsUISupported = 21,
DISPID_SVDisplayUI = 22
} DISPID_SpeechVoice;
typedef enum {
DISPID_SVSCurrentStreamNumber = 1,
DISPID_SVSLastStreamNumberQueued = 2,
DISPID_SVSLastResult = 3,
DISPID_SVSRunningState = 4,
DISPID_SVSInputWordPosition = 5,
DISPID_SVSInputWordLength = 6,
DISPID_SVSInputSentencePosition = 7,
DISPID_SVSInputSentenceLength = 8,
DISPID_SVSLastBookmark = 9,
DISPID_SVSLastBookmarkId = 10,
DISPID_SVSPhonemeId = 11,
DISPID_SVSVisemeId = 12
} DISPID_SpeechVoiceStatus;
typedef enum {
DISPID_SVEStreamStart = 1,
DISPID_SVEStreamEnd = 2,
DISPID_SVEVoiceChange = 3,
DISPID_SVEBookmark = 4,
DISPID_SVEWord = 5,
DISPID_SVEPhoneme = 6,
DISPID_SVESentenceBoundary = 7,
DISPID_SVEViseme = 8,
DISPID_SVEAudioLevel = 9,
DISPID_SVEEnginePrivate = 10
} DISPID_SpeechVoiceEvent;
typedef enum {
DISPID_SRRecognizer = 1,
DISPID_SRAllowAudioInputFormatChangesOnNextSet = 2,
DISPID_SRAudioInput = 3,
DISPID_SRAudioInputStream = 4,
DISPID_SRIsShared = 5,
DISPID_SRState = 6,
DISPID_SRStatus = 7,
DISPID_SRProfile = 8,
DISPID_SREmulateRecognition = 9,
DISPID_SRCreateRecoContext = 10,
DISPID_SRGetFormat = 11,
DISPID_SRSetPropertyNumber = 12,
DISPID_SRGetPropertyNumber = 13,
DISPID_SRSetPropertyString = 14,
DISPID_SRGetPropertyString = 15,
DISPID_SRIsUISupported = 16,
DISPID_SRDisplayUI = 17,
DISPID_SRGetRecognizers = 18,
DISPID_SVGetAudioInputs = 19,
DISPID_SVGetProfiles = 20
} DISPID_SpeechRecognizer;
typedef enum {
SECFIgnoreCase = 1,
SECFIgnoreKanaType = 0x00010000,
SECFIgnoreWidth = 0x00020000,
SECFNoSpecialChars = 0x20000000,
SECFEmulateResult = 0x40000000,
SECFDefault = 0x00030001
} SpeechEmulationCompareFlags;
typedef enum {
DISPID_SRSAudioStatus = 1,
DISPID_SRSCurrentStreamPosition = 2,
DISPID_SRSCurrentStreamNumber = 3,
DISPID_SRSNumberOfActiveRules = 4,
DISPID_SRSClsidEngine = 5,
DISPID_SRSSupportedLanguages = 6
} DISPID_SpeechRecognizerStatus;
typedef enum {
DISPID_SRCRecognizer = 1,
DISPID_SRCAudioInInterferenceStatus = 2,
DISPID_SRCRequestedUIType = 3,
DISPID_SRCVoice = 4,
DISPID_SRAllowVoiceFormatMatchingOnNextSet = 5,
DISPID_SRCVoicePurgeEvent = 6,
DISPID_SRCEventInterests = 7,
DISPID_SRCCmdMaxAlternates = 8,
DISPID_SRCState = 9,
DISPID_SRCRetainedAudio = 10,
DISPID_SRCRetainedAudioFormat = 11,
DISPID_SRCPause = 12,
DISPID_SRCResume = 13,
DISPID_SRCCreateGrammar = 14,
DISPID_SRCCreateResultFromMemory = 15,
DISPID_SRCBookmark = 16,
DISPID_SRCSetAdaptationData = 17
} DISPID_SpeechRecoContext;
typedef enum {
DISPID_SRGId = 1,
DISPID_SRGRecoContext = 2,
DISPID_SRGState = 3,
DISPID_SRGRules = 4,
DISPID_SRGReset = 5,
DISPID_SRGCommit = 6,
DISPID_SRGCmdLoadFromFile = 7,
DISPID_SRGCmdLoadFromObject = 8,
DISPID_SRGCmdLoadFromResource = 9,
DISPID_SRGCmdLoadFromMemory = 10,
DISPID_SRGCmdLoadFromProprietaryGrammar = 11,
DISPID_SRGCmdSetRuleState = 12,
DISPID_SRGCmdSetRuleIdState = 13,
DISPID_SRGDictationLoad = 14,
DISPID_SRGDictationUnload = 15,
DISPID_SRGDictationSetState = 16,
DISPID_SRGSetWordSequenceData = 17,
DISPID_SRGSetTextSelection = 18,
DISPID_SRGIsPronounceable = 19
} DISPIDSPRG;
typedef enum {
DISPID_SRCEStartStream = 1,
DISPID_SRCEEndStream = 2,
DISPID_SRCEBookmark = 3,
DISPID_SRCESoundStart = 4,
DISPID_SRCESoundEnd = 5,
DISPID_SRCEPhraseStart = 6,
DISPID_SRCERecognition = 7,
DISPID_SRCEHypothesis = 8,
DISPID_SRCEPropertyNumberChange = 9,
DISPID_SRCEPropertyStringChange = 10,
DISPID_SRCEFalseRecognition = 11,
DISPID_SRCEInterference = 12,
DISPID_SRCERequestUI = 13,
DISPID_SRCERecognizerStateChange = 14,
DISPID_SRCEAdaptation = 15,
DISPID_SRCERecognitionForOtherContext = 16,
DISPID_SRCEAudioLevel = 17,
DISPID_SRCEEnginePrivate = 18
} DISPID_SpeechRecoContextEvents;
typedef enum {
DISPID_SGRAttributes = 1,
DISPID_SGRInitialState = 2,
DISPID_SGRName = 3,
DISPID_SGRId = 4,
DISPID_SGRClear = 5,
DISPID_SGRAddResource = 6,
DISPID_SGRAddState = 7
} DISPID_SpeechGrammarRule;
typedef enum {
DISPID_SGRsCount = 1,
DISPID_SGRsDynamic = 2,
DISPID_SGRsAdd = 3,
DISPID_SGRsCommit = 4,
DISPID_SGRsCommitAndSave = 5,
DISPID_SGRsFindRule = 6,
DISPID_SGRsItem = 0,
DISPID_SGRs_NewEnum = 0xfffffffc
} DISPID_SpeechGrammarRules;
typedef enum {
DISPID_SGRSRule = 1,
DISPID_SGRSTransitions = 2,
DISPID_SGRSAddWordTransition = 3,
DISPID_SGRSAddRuleTransition = 4,
DISPID_SGRSAddSpecialTransition = 5
} DISPID_SpeechGrammarRuleState;
typedef enum {
DISPID_SGRSTsCount = 1,
DISPID_SGRSTsItem = 0,
DISPID_SGRSTs_NewEnum = 0xfffffffc
} DISPID_SpeechGrammarRuleStateTransitions;
typedef enum {
DISPID_SGRSTType = 1,
DISPID_SGRSTText = 2,
DISPID_SGRSTRule = 3,
DISPID_SGRSTWeight = 4,
DISPID_SGRSTPropertyName = 5,
DISPID_SGRSTPropertyId = 6,
DISPID_SGRSTPropertyValue = 7,
DISPID_SGRSTNextState = 8
} DISPID_SpeechGrammarRuleStateTransition;
typedef enum {
DISPIDSPTSI_ActiveOffset = 1,
DISPIDSPTSI_ActiveLength = 2,
DISPIDSPTSI_SelectionOffset = 3,
DISPIDSPTSI_SelectionLength = 4
} DISPIDSPTSI;
typedef enum {
DISPID_SRRRecoContext = 1,
DISPID_SRRTimes = 2,
DISPID_SRRAudioFormat = 3,
DISPID_SRRPhraseInfo = 4,
DISPID_SRRAlternates = 5,
DISPID_SRRAudio = 6,
DISPID_SRRSpeakAudio = 7,
DISPID_SRRSaveToMemory = 8,
DISPID_SRRDiscardResultInfo = 9
} DISPID_SpeechRecoResult;
typedef enum {
DISPID_SRRGetXMLResult = 10,
DISPID_SRRGetXMLErrorInfo = 11
} DISPID_SpeechXMLRecoResult;
[
odl,
uuid(AAEC54AF-8F85-4924-944D-B79D39D72E19),
helpstring("ISpeechXMLRecoResult Interface"),
dual,
oleautomation
]
interface ISpeechXMLRecoResult : ISpeechRecoResult {
[id(0x0000000a), helpstring("GetXMLResult")]
HRESULT GetXMLResult(
[in] SPXMLRESULTOPTIONS Options,
[out, retval] BSTR* pResult);
[id(0x0000000b), helpstring("GetXMLErrorInfo")]
HRESULT GetXMLErrorInfo(
[out] long* LineNumber,
[out] BSTR* ScriptLine,
[out] BSTR* Source,
[out] BSTR* Description,
[out] long* ResultCode,
[out, retval] VARIANT_BOOL* IsError);
};
typedef enum {
SPXRO_SML = 0,
SPXRO_Alternates_SML = 1
} SPXMLRESULTOPTIONS;
typedef enum {
DISPID_SRRSetTextFeedback = 12
} DISPID_SpeechRecoResult2;
[
odl,
uuid(6D60EB64-ACED-40A6-BBF3-4E557F71DEE2),
helpstring("ISpeechRecoResultDispatch Interface"),
hidden,
dual,
oleautomation
]
interface ISpeechRecoResultDispatch : IDispatch {
[id(0x00000001), propget, helpstring("RecoContext")]
HRESULT RecoContext([out, retval] ISpeechRecoContext** RecoContext);
[id(0x00000002), propget, helpstring("Times")]
HRESULT Times([out, retval] ISpeechRecoResultTimes** Times);
[id(0x00000003), propputref, helpstring("AudioFormat")]
HRESULT AudioFormat([in] ISpeechAudioFormat* Format);
[id(0x00000003), propget, helpstring("AudioFormat")]
HRESULT AudioFormat([out, retval] ISpeechAudioFormat** Format);
[id(0x00000004), propget, helpstring("PhraseInfo")]
HRESULT PhraseInfo([out, retval] ISpeechPhraseInfo** PhraseInfo);
[id(0x00000005), helpstring("Alternates")]
HRESULT Alternates(
[in] long RequestCount,
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[out, retval] ISpeechPhraseAlternates** Alternates);
[id(0x00000006), helpstring("Audio")]
HRESULT Audio(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[out, retval] ISpeechMemoryStream** Stream);
[id(0x00000007), helpstring("SpeakAudio")]
HRESULT SpeakAudio(
[in, optional, defaultvalue(0)] long StartElement,
[in, optional, defaultvalue(-1)] long Elements,
[in, optional, defaultvalue(0)] SpeechVoiceSpeakFlags Flags,
[out, retval] long* StreamNumber);
[id(0x00000008), helpstring("SaveToMemory")]
HRESULT SaveToMemory([out, retval] VARIANT* ResultBlock);
[id(0x00000009), helpstring("DiscardResultInfo")]
HRESULT DiscardResultInfo([in] SpeechDiscardType ValueTypes);
[id(0x0000000a), helpstring("GetXMLResult")]
HRESULT GetXMLResult(
[in] SPXMLRESULTOPTIONS Options,
[out, retval] BSTR* pResult);
[id(0x0000000b), helpstring("GetXMLErrorInfo")]
HRESULT GetXMLErrorInfo(
[out] long* LineNumber,
[out] BSTR* ScriptLine,
[out] BSTR* Source,
[out] BSTR* Description,
[out] HRESULT* ResultCode,
[out, retval] VARIANT_BOOL* IsError);
[id(0x0000000c), helpstring("SetTextFeedback")]
HRESULT SetTextFeedback(
[in] BSTR Feedback,
[in] VARIANT_BOOL WasSuccessful);
};
typedef enum {
DISPID_SPPBRestorePhraseFromMemory = 1
} DISPID_SpeechPhraseBuilder;
[
odl,
uuid(3B151836-DF3A-4E0A-846C-D2ADC9334333),
helpstring("ISpeechPhraseInfoBuilder Interface"),
dual,
oleautomation
]
interface ISpeechPhraseInfoBuilder : IDispatch {
[id(0x00000001), helpstring("RestorePhraseFromMemory")]
HRESULT RestorePhraseFromMemory(
[in] VARIANT* PhraseInMemory,
[out, retval] ISpeechPhraseInfo** PhraseInfo);
};
typedef enum {
DISPID_SRRTStreamTime = 1,
DISPID_SRRTLength = 2,
DISPID_SRRTTickCount = 3,
DISPID_SRRTOffsetFromStart = 4
} DISPID_SpeechRecoResultTimes;
typedef enum {
DISPID_SPARecoResult = 1,
DISPID_SPAStartElementInResult = 2,
DISPID_SPANumberOfElementsInResult = 3,
DISPID_SPAPhraseInfo = 4,
DISPID_SPACommit = 5
} DISPID_SpeechPhraseAlternate;
typedef enum {
DISPID_SPAsCount = 1,
DISPID_SPAsItem = 0,
DISPID_SPAs_NewEnum = 0xfffffffc
} DISPID_SpeechPhraseAlternates;
typedef enum {
DISPID_SPILanguageId = 1,
DISPID_SPIGrammarId = 2,
DISPID_SPIStartTime = 3,
DISPID_SPIAudioStreamPosition = 4,
DISPID_SPIAudioSizeBytes = 5,
DISPID_SPIRetainedSizeBytes = 6,
DISPID_SPIAudioSizeTime = 7,
DISPID_SPIRule = 8,
DISPID_SPIProperties = 9,
DISPID_SPIElements = 10,
DISPID_SPIReplacements = 11,
DISPID_SPIEngineId = 12,
DISPID_SPIEnginePrivateData = 13,
DISPID_SPISaveToMemory = 14,
DISPID_SPIGetText = 15,
DISPID_SPIGetDisplayAttributes = 16
} DISPID_SpeechPhraseInfo;
typedef enum {
DISPID_SPEAudioTimeOffset = 1,
DISPID_SPEAudioSizeTime = 2,
DISPID_SPEAudioStreamOffset = 3,
DISPID_SPEAudioSizeBytes = 4,
DISPID_SPERetainedStreamOffset = 5,
DISPID_SPERetainedSizeBytes = 6,
DISPID_SPEDisplayText = 7,
DISPID_SPELexicalForm = 8,
DISPID_SPEPronunciation = 9,
DISPID_SPEDisplayAttributes = 10,
DISPID_SPERequiredConfidence = 11,
DISPID_SPEActualConfidence = 12,
DISPID_SPEEngineConfidence = 13
} DISPID_SpeechPhraseElement;
typedef enum {
DISPID_SPEsCount = 1,
DISPID_SPEsItem = 0,
DISPID_SPEs_NewEnum = 0xfffffffc
} DISPID_SpeechPhraseElements;
typedef enum {
DISPID_SPRDisplayAttributes = 1,
DISPID_SPRText = 2,
DISPID_SPRFirstElement = 3,
DISPID_SPRNumberOfElements = 4
} DISPID_SpeechPhraseReplacement;
typedef enum {
DISPID_SPRsCount = 1,
DISPID_SPRsItem = 0,
DISPID_SPRs_NewEnum = 0xfffffffc
} DISPID_SpeechPhraseReplacements;
typedef enum {
DISPID_SPPName = 1,
DISPID_SPPId = 2,
DISPID_SPPValue = 3,
DISPID_SPPFirstElement = 4,
DISPID_SPPNumberOfElements = 5,
DISPID_SPPEngineConfidence = 6,
DISPID_SPPConfidence = 7,
DISPID_SPPParent = 8,
DISPID_SPPChildren = 9
} DISPID_SpeechPhraseProperty;
typedef enum {
DISPID_SPPsCount = 1,
DISPID_SPPsItem = 0,
DISPID_SPPs_NewEnum = 0xfffffffc
} DISPID_SpeechPhraseProperties;
typedef enum {
DISPID_SPRuleName = 1,
DISPID_SPRuleId = 2,
DISPID_SPRuleFirstElement = 3,
DISPID_SPRuleNumberOfElements = 4,
DISPID_SPRuleParent = 5,
DISPID_SPRuleChildren = 6,
DISPID_SPRuleConfidence = 7,
DISPID_SPRuleEngineConfidence = 8
} DISPID_SpeechPhraseRule;
typedef enum {
DISPID_SPRulesCount = 1,
DISPID_SPRulesItem = 0,
DISPID_SPRules_NewEnum = 0xfffffffc
} DISPID_SpeechPhraseRules;
typedef enum {
DISPID_SLGenerationId = 1,
DISPID_SLGetWords = 2,
DISPID_SLAddPronunciation = 3,
DISPID_SLAddPronunciationByPhoneIds = 4,
DISPID_SLRemovePronunciation = 5,
DISPID_SLRemovePronunciationByPhoneIds = 6,
DISPID_SLGetPronunciations = 7,
DISPID_SLGetGenerationChange = 8
} DISPID_SpeechLexicon;
typedef enum {
DISPID_SLWsCount = 1,
DISPID_SLWsItem = 0,
DISPID_SLWs_NewEnum = 0xfffffffc
} DISPID_SpeechLexiconWords;
typedef enum {
DISPID_SLWLangId = 1,
DISPID_SLWType = 2,
DISPID_SLWWord = 3,
DISPID_SLWPronunciations = 4
} DISPID_SpeechLexiconWord;
typedef enum {
DISPID_SLPsCount = 1,
DISPID_SLPsItem = 0,
DISPID_SLPs_NewEnum = 0xfffffffc
} DISPID_SpeechLexiconProns;
typedef enum {
DISPID_SLPType = 1,
DISPID_SLPLangId = 2,
DISPID_SLPPartOfSpeech = 3,
DISPID_SLPPhoneIds = 4,
DISPID_SLPSymbolic = 5
} DISPID_SpeechLexiconPronunciation;
typedef enum {
DISPID_SPCLangId = 1,
DISPID_SPCPhoneToId = 2,
DISPID_SPCIdToPhone = 3
} DISPID_SpeechPhoneConverter;
[
odl,
uuid(C3E4F353-433F-43D6-89A1-6A62A7054C3D),
helpstring("ISpeechPhoneConverter Interface"),
dual,
oleautomation
]
interface ISpeechPhoneConverter : IDispatch {
[id(0x00000001), propget, helpstring("LanguageId")]
HRESULT LanguageId([out, retval] long* LanguageId);
[id(0x00000001), propput, helpstring("LanguageId")]
HRESULT LanguageId([in] long LanguageId);
[id(0x00000002), helpstring("PhoneToId")]
HRESULT PhoneToId(
[in] BSTR Phonemes,
[out, retval] VARIANT* IdArray);
[id(0x00000003), helpstring("IdToPhone")]
HRESULT IdToPhone(
[in] VARIANT IdArray,
[out, retval] BSTR* Phonemes);
};
[
uuid(E2AE5372-5D40-11D2-960E-00C04F8EE628),
helpstring("SpNotify"),
hidden,
restricted
]
coclass SpNotifyTranslator {
[default] interface ISpNotifyTranslator;
};
[
odl,
uuid(ACA16614-5D3D-11D2-960E-00C04F8EE628),
helpstring("ISpNotifyTranslator Interface"),
restricted
]
interface ISpNotifyTranslator : ISpNotifySink {
HRESULT _stdcall InitWindowMessage(
[in] wireHWND hWnd,
[in] unsigned int Msg,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall InitCallback(
[in] void** pfnCallback,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall InitSpNotifyCallback(
[in] void** pSpCallback,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall InitWin32Event(
[in] void* hEvent,
[in] long fCloseHandleOnRelease);
HRESULT _stdcall Wait([in] unsigned long dwMilliseconds);
void* _stdcall GetEventHandle();
};
[
odl,
uuid(259684DC-37C3-11D2-9603-00C04F8EE628),
helpstring("ISpNotifySink Interface"),
restricted
]
interface ISpNotifySink : IUnknown {
HRESULT _stdcall Notify();
};
typedef [public]
_RemotableHandle* wireHWND;
typedef struct tag_RemotableHandle {
long fContext;
__MIDL_IWinTypes_0009 u;
} _RemotableHandle;
typedef union tag__MIDL_IWinTypes_0009 {
long hInproc;
long hRemote;
} __MIDL_IWinTypes_0009;
typedef [public ,
custom(F914481D-9C62-4B43-9340-E9B2E6252E5F, 1)
]
uint64 UINT_PTR;
typedef [public ,
custom(F914481D-9C62-4B43-9340-E9B2E6252E5F, 1)
]
int64 LONG_PTR;
[
uuid(A910187F-0C7A-45AC-92CC-59EDAFB77B53),
helpstring("SpObjectTokenCategory Class")
]
coclass SpObjectTokenCategory {
[default] interface ISpeechObjectTokenCategory;
interface ISpObjectTokenCategory;
};
[
odl,
uuid(2D3D3845-39AF-4850-BBF9-40B49780011D),
helpstring("ISpObjectTokenCategory"),
restricted
]
interface ISpObjectTokenCategory : ISpDataKey {
HRESULT _stdcall SetId(
[in] LPWSTR pszCategoryId,
[in] long fCreateIfNotExist);
HRESULT _stdcall GetId([out] LPWSTR* ppszCoMemCategoryId);
HRESULT _stdcall GetDataKey(
[in] SPDATAKEYLOCATION spdkl,
[out] ISpDataKey** ppDataKey);
HRESULT _stdcall EnumTokens(
[in] LPWSTR pzsReqAttribs,
[in] LPWSTR pszOptAttribs,
[out] IEnumSpObjectTokens** ppEnum);
HRESULT _stdcall SetDefaultTokenId([in] LPWSTR pszTokenId);
HRESULT _stdcall GetDefaultTokenId([out] LPWSTR* ppszCoMemTokenId);
};
[
odl,
uuid(14056581-E16C-11D2-BB90-00C04F8EE6C0),
helpstring("ISpDataKey Interface"),
restricted
]
interface ISpDataKey : IUnknown {
HRESULT _stdcall SetData(
[in] LPWSTR pszValueName,
[in] unsigned long cbData,
[in] unsigned char* pData);
HRESULT _stdcall GetData(
[in] LPWSTR pszValueName,
[in] unsigned long* pcbData,
[out] unsigned char* pData);
HRESULT _stdcall SetStringValue(
[in] LPWSTR pszValueName,
[in] LPWSTR pszValue);
HRESULT _stdcall GetStringValue(
[in] LPWSTR pszValueName,
[out] LPWSTR* ppszValue);
HRESULT _stdcall SetDWORD(
[in] LPWSTR pszValueName,
[in] unsigned long dwValue);
HRESULT _stdcall GetDWORD(
[in] LPWSTR pszValueName,
[out] unsigned long* pdwValue);
HRESULT _stdcall OpenKey(
[in] LPWSTR pszSubKeyName,
[out] ISpDataKey** ppSubKey);
HRESULT _stdcall CreateKey(
[in] LPWSTR pszSubKey,
[out] ISpDataKey** ppSubKey);
HRESULT _stdcall DeleteKey([in] LPWSTR pszSubKey);
HRESULT _stdcall DeleteValue([in] LPWSTR pszValueName);
HRESULT _stdcall EnumKeys(
[in] unsigned long Index,
[out] LPWSTR* ppszSubKeyName);
HRESULT _stdcall EnumValues(
[in] unsigned long Index,
[out] LPWSTR* ppszValueName);
};
typedef enum {
SPDKL_DefaultLocation = 0,
SPDKL_CurrentUser = 1,
SPDKL_LocalMachine = 2,
SPDKL_CurrentConfig = 5
} SPDATAKEYLOCATION;
[
odl,
uuid(06B64F9E-7FDA-11D2-B4F2-00C04F797396),
helpstring("IEnumSpObjectTokens Interface"),
restricted
]
interface IEnumSpObjectTokens : IUnknown {
HRESULT _stdcall Next(
[in] unsigned long celt,
[out] ISpObjectToken** pelt,
[out] unsigned long* pceltFetched);
HRESULT _stdcall Skip([in] unsigned long celt);
HRESULT _stdcall Reset();
HRESULT _stdcall Clone([out] IEnumSpObjectTokens** ppEnum);
HRESULT _stdcall Item(
[in] unsigned long Index,
[out] ISpObjectToken** ppToken);
HRESULT _stdcall GetCount([out] unsigned long* pCount);
};
[
odl,
uuid(14056589-E16C-11D2-BB90-00C04F8EE6C0),
helpstring("ISpObjectToken Interface"),
restricted
]
interface ISpObjectToken : ISpDataKey {
HRESULT _stdcall SetId(
LPWSTR pszCategoryId,
[in] LPWSTR pszTokenId,
[in] long fCreateIfNotExist);
HRESULT _stdcall GetId([out] LPWSTR* ppszCoMemTokenId);
HRESULT _stdcall GetCategory([out] ISpObjectTokenCategory** ppTokenCategory);
HRESULT _stdcall CreateInstance(
[in] IUnknown* pUnkOuter,
[in] unsigned long dwClsContext,
[in] GUID* riid,
[out] void** ppvObject);
HRESULT _stdcall GetStorageFileName(
[in] GUID* clsidCaller,
[in] LPWSTR pszValueName,
[in] LPWSTR pszFileNameSpecifier,
[in] unsigned long nFolder,
[out] LPWSTR* ppszFilePath);
HRESULT _stdcall RemoveStorageFileName(
[in] GUID* clsidCaller,
[in] LPWSTR pszKeyName,
[in] long fDeleteFile);
HRESULT _stdcall Remove(GUID* pclsidCaller);
HRESULT _stdcall IsUISupported(
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData,
[in] IUnknown* punkObject,
[out] long* pfSupported);
HRESULT _stdcall DisplayUI(
[in] wireHWND hWndParent,
[in] LPWSTR pszTitle,
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData,
[in] IUnknown* punkObject);
HRESULT _stdcall MatchesAttributes(
[in] LPWSTR pszAttributes,
[out] long* pfMatches);
};
[
uuid(EF411752-3736-4CB4-9C8C-8EF4CCB58EFE),
helpstring("SpObjectToken Class")
]
coclass SpObjectToken {
[default] interface ISpeechObjectToken;
interface ISpObjectToken;
};
[
uuid(96749373-3391-11D2-9EE3-00C04F797396),
helpstring("SpResourceManger"),
hidden,
restricted
]
coclass SpResourceManager {
[default] interface ISpResourceManager;
};
[
odl,
uuid(93384E18-5014-43D5-ADBB-A78E055926BD),
helpstring("ISpResourceManager Interface"),
restricted
]
interface ISpResourceManager : IServiceProvider {
HRESULT _stdcall SetObject(
[in] GUID* guidServiceId,
[in] IUnknown* punkObject);
HRESULT _stdcall GetObject(
[in] GUID* guidServiceId,
[in] GUID* ObjectCLSID,
[in] GUID* ObjectIID,
[in] long fReleaseWhenLastExternalRefReleased,
[out] void** ppObject);
};
[
odl,
uuid(6D5140C1-7436-11CE-8034-00AA006009FA)
]
interface IServiceProvider : IUnknown {
HRESULT _stdcall RemoteQueryService(
[in] GUID* guidService,
[in] GUID* riid,
[out] IUnknown** ppvObject);
};
[
uuid(7013943A-E2EC-11D2-A086-00C04F8EF9B5),
helpstring("FormatConverter Class"),
hidden,
restricted
]
coclass SpStreamFormatConverter {
[default] interface ISpStreamFormatConverter;
};
[
odl,
uuid(678A932C-EA71-4446-9B41-78FDA6280A29),
helpstring("ISpStreamFormatConverter Interface"),
restricted
]
interface ISpStreamFormatConverter : ISpStreamFormat {
HRESULT _stdcall SetBaseStream(
[in] ISpStreamFormat* pStream,
[in] long fSetFormatToBaseStreamFormat,
[in] long fWriteToBaseStream);
HRESULT _stdcall GetBaseStream([out] ISpStreamFormat** ppStream);
HRESULT _stdcall SetFormat(
[in] GUID* rguidFormatIdOfConvertedStream,
[in] WAVEFORMATEX* pWaveFormatExOfConvertedStream);
HRESULT _stdcall ResetSeekPosition();
HRESULT _stdcall ScaleConvertedToBaseOffset(
[in] uint64 ullOffsetConvertedStream,
[out] uint64* pullOffsetBaseStream);
HRESULT _stdcall ScaleBaseToConvertedOffset(
[in] uint64 ullOffsetBaseStream,
[out] uint64* pullOffsetConvertedStream);
};
[
odl,
uuid(BED530BE-2606-4F4D-A1C0-54C5CDA5566F),
helpstring("ISpStreamFormat Interface"),
restricted
]
interface ISpStreamFormat : IStream {
HRESULT _stdcall GetFormat(
[in] GUID* pguidFormatId,
[out] WAVEFORMATEX** ppCoMemWaveFormatEx);
};
[
odl,
uuid(0000000C-0000-0000-C000-000000000046)
]
interface IStream : ISequentialStream {
HRESULT _stdcall RemoteSeek(
[in] _LARGE_INTEGER dlibMove,
[in] unsigned long dwOrigin,
[out] _ULARGE_INTEGER* plibNewPosition);
HRESULT _stdcall SetSize([in] _ULARGE_INTEGER libNewSize);
HRESULT _stdcall RemoteCopyTo(
[in] IStream* pstm,
[in] _ULARGE_INTEGER cb,
[out] _ULARGE_INTEGER* pcbRead,
[out] _ULARGE_INTEGER* pcbWritten);
HRESULT _stdcall Commit([in] unsigned long grfCommitFlags);
HRESULT _stdcall Revert();
HRESULT _stdcall LockRegion(
[in] _ULARGE_INTEGER libOffset,
[in] _ULARGE_INTEGER cb,
[in] unsigned long dwLockType);
HRESULT _stdcall UnlockRegion(
[in] _ULARGE_INTEGER libOffset,
[in] _ULARGE_INTEGER cb,
[in] unsigned long dwLockType);
HRESULT _stdcall Stat(
[out] tagSTATSTG* pstatstg,
[in] unsigned long grfStatFlag);
HRESULT _stdcall Clone([out] IStream** ppstm);
};
[
odl,
uuid(0C733A30-2A1C-11CE-ADE5-00AA0044773D)
]
interface ISequentialStream : IUnknown {
HRESULT _stdcall RemoteRead(
[out] unsigned char* pv,
[in] unsigned long cb,
[out] unsigned long* pcbRead);
HRESULT _stdcall RemoteWrite(
[in] unsigned char* pv,
[in] unsigned long cb,
[out] unsigned long* pcbWritten);
};
typedef struct tag_LARGE_INTEGER {
int64 QuadPart;
} _LARGE_INTEGER;
typedef struct tag_ULARGE_INTEGER {
uint64 QuadPart;
} _ULARGE_INTEGER;
typedef struct tagtagSTATSTG {
LPWSTR pwcsName;
unsigned long Type;
_ULARGE_INTEGER cbSize;
_FILETIME mtime;
_FILETIME ctime;
_FILETIME atime;
unsigned long grfMode;
unsigned long grfLocksSupported;
GUID clsid;
unsigned long grfStateBits;
unsigned long reserved;
} tagSTATSTG;
typedef struct tag_FILETIME {
unsigned long dwLowDateTime;
unsigned long dwHighDateTime;
} _FILETIME;
typedef struct tagWAVEFORMATEX {
unsigned short wFormatTag;
unsigned short nChannels;
unsigned long nSamplesPerSec;
unsigned long nAvgBytesPerSec;
unsigned short nBlockAlign;
unsigned short wBitsPerSample;
unsigned short cbSize;
} WAVEFORMATEX;
[
uuid(AB1890A0-E91F-11D2-BB91-00C04F8EE6C0),
helpstring("SpMMAudioEnum Class"),
hidden,
restricted
]
coclass SpMMAudioEnum {
[default] interface IEnumSpObjectTokens;
};
[
uuid(CF3D2E50-53F2-11D2-960C-00C04F8EE628),
helpstring("SpMMAudioIn Class")
]
coclass SpMMAudioIn {
[default] interface ISpeechMMSysAudio;
interface ISpEventSource;
interface ISpEventSink;
interface ISpObjectWithToken;
interface ISpMMSysAudio;
};
[
odl,
uuid(BE7A9CCE-5F9E-11D2-960F-00C04F8EE628),
helpstring("ISpEventSource Interface"),
restricted
]
interface ISpEventSource : ISpNotifySource {
HRESULT _stdcall SetInterest(
[in] uint64 ullEventInterest,
[in] uint64 ullQueuedInterest);
HRESULT _stdcall GetEvents(
[in] unsigned long ulCount,
[out] SPEVENT* pEventArray,
[out] unsigned long* pulFetched);
HRESULT _stdcall GetInfo([out] SPEVENTSOURCEINFO* pInfo);
};
[
odl,
uuid(5EFF4AEF-8487-11D2-961C-00C04F8EE628),
helpstring("ISpNotifySource Interface"),
restricted
]
interface ISpNotifySource : IUnknown {
HRESULT _stdcall SetNotifySink([in] ISpNotifySink* pNotifySink);
HRESULT _stdcall SetNotifyWindowMessage(
[in] wireHWND hWnd,
[in] unsigned int Msg,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall SetNotifyCallbackFunction(
[in] void** pfnCallback,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall SetNotifyCallbackInterface(
[in] void** pSpCallback,
[in] UINT_PTR wParam,
[in] LONG_PTR lParam);
HRESULT _stdcall SetNotifyWin32Event();
HRESULT _stdcall WaitForNotifyEvent([in] unsigned long dwMilliseconds);
void* _stdcall GetNotifyEventHandle();
};
typedef struct tagSPEVENT {
unsigned short eEventId;
unsigned short elParamType;
unsigned long ulStreamNum;
uint64 ullAudioStreamOffset;
UINT_PTR wParam;
LONG_PTR lParam;
} SPEVENT;
typedef struct tagSPEVENTSOURCEINFO {
uint64 ullEventInterest;
uint64 ullQueuedInterest;
unsigned long ulCount;
} SPEVENTSOURCEINFO;
[
odl,
uuid(BE7A9CC9-5F9E-11D2-960F-00C04F8EE628),
helpstring("ISpEventSink Interface"),
restricted
]
interface ISpEventSink : IUnknown {
HRESULT _stdcall AddEvents(
[in] SPEVENT* pEventArray,
[in] unsigned long ulCount);
HRESULT _stdcall GetEventInterest([out] uint64* pullEventInterest);
};
[
odl,
uuid(5B559F40-E952-11D2-BB91-00C04F8EE6C0),
helpstring("ISpObjectWithToken Interface"),
restricted
]
interface ISpObjectWithToken : IUnknown {
HRESULT _stdcall SetObjectToken([in] ISpObjectToken* pToken);
HRESULT _stdcall GetObjectToken([out] ISpObjectToken** ppToken);
};
[
odl,
uuid(15806F6E-1D70-4B48-98E6-3B1A007509AB),
helpstring("ISpMMSysAudio Interface"),
restricted
]
interface ISpMMSysAudio : ISpAudio {
HRESULT _stdcall GetDeviceId([out] unsigned int* puDeviceId);
HRESULT _stdcall SetDeviceId([in] unsigned int uDeviceId);
HRESULT _stdcall GetMMHandle([out] void** pHandle);
HRESULT _stdcall GetLineId([out] unsigned int* puLineId);
HRESULT _stdcall SetLineId([in] unsigned int uLineId);
};
[
odl,
uuid(C05C768F-FAE8-4EC2-8E07-338321C12452),
helpstring("ISpAudio Interface"),
restricted
]
interface ISpAudio : ISpStreamFormat {
HRESULT _stdcall SetState(
[in] SPAUDIOSTATE NewState,
[in] uint64 ullReserved);
HRESULT _stdcall SetFormat(
[in] GUID* rguidFmtId,
[in] WAVEFORMATEX* pWaveFormatEx);
HRESULT _stdcall GetStatus([out] SPAUDIOSTATUS* pStatus);
HRESULT _stdcall SetBufferInfo([in] SPAUDIOBUFFERINFO* pBuffInfo);
HRESULT _stdcall GetBufferInfo([out] SPAUDIOBUFFERINFO* pBuffInfo);
HRESULT _stdcall GetDefaultFormat(
[out] GUID* pFormatId,
[out] WAVEFORMATEX** ppCoMemWaveFormatEx);
void* _stdcall EventHandle();
HRESULT _stdcall GetVolumeLevel([out] unsigned long* pLevel);
HRESULT _stdcall SetVolumeLevel([in] unsigned long Level);
HRESULT _stdcall GetBufferNotifySize([out] unsigned long* pcbSize);
HRESULT _stdcall SetBufferNotifySize([in] unsigned long cbSize);
};
typedef [public]
_SPAUDIOSTATE SPAUDIOSTATE;
typedef enum {
SPAS_CLOSED = 0,
SPAS_STOP = 1,
SPAS_PAUSE = 2,
SPAS_RUN = 3
} _SPAUDIOSTATE;
typedef struct tagSPAUDIOSTATUS {
long cbFreeBuffSpace;
unsigned long cbNonBlockingIO;
SPAUDIOSTATE State;
uint64 CurSeekPos;
uint64 CurDevicePos;
unsigned long dwAudioLevel;
unsigned long dwReserved2;
} SPAUDIOSTATUS;
typedef struct tagSPAUDIOBUFFERINFO {
unsigned long ulMsMinNotification;
unsigned long ulMsBufferSize;
unsigned long ulMsEventBias;
} SPAUDIOBUFFERINFO;
[
uuid(A8C680EB-3D32-11D2-9EE7-00C04F797396),
helpstring("SpMMAudioOut Class")
]
coclass SpMMAudioOut {
[default] interface ISpeechMMSysAudio;
interface ISpEventSource;
interface ISpEventSink;
interface ISpObjectWithToken;
interface ISpMMSysAudio;
};
[
uuid(715D9C59-4442-11D2-9605-00C04F8EE628),
helpstring("SpStream Class"),
hidden,
restricted
]
coclass SpStream {
[default] interface ISpStream;
};
[
odl,
uuid(12E3CCA9-7518-44C5-A5E7-BA5A79CB929E),
helpstring("ISpStream Interface"),
restricted
]
interface ISpStream : ISpStreamFormat {
HRESULT _stdcall SetBaseStream(
[in] IStream* pStream,
[in] GUID* rguidFormat,
[in] WAVEFORMATEX* pWaveFormatEx);
HRESULT _stdcall GetBaseStream([out] IStream** ppStream);
HRESULT _stdcall BindToFile(
[in] LPWSTR pszFileName,
[in] SPFILEMODE eMode,
[in] GUID* pFormatId,
WAVEFORMATEX* pWaveFormatEx,
[in] uint64 ullEventInterest);
HRESULT _stdcall Close();
};
typedef enum {
SPFM_OPEN_READONLY = 0,
SPFM_OPEN_READWRITE = 1,
SPFM_CREATE = 2,
SPFM_CREATE_ALWAYS = 3,
SPFM_NUM_MODES = 4
} SPFILEMODE;
[
uuid(96749377-3391-11D2-9EE3-00C04F797396),
helpstring("SpVoice Class")
]
coclass SpVoice {
[default] interface ISpeechVoice;
interface ISpVoice;
interface ISpPhoneticAlphabetSelection;
[default, source] dispinterface _ISpeechVoiceEvents;
};
[
odl,
uuid(6C44DF74-72B9-4992-A1EC-EF996E0422D4),
helpstring("ISpVoice Interface"),
restricted
]
interface ISpVoice : ISpEventSource {
HRESULT _stdcall SetOutput(
[in] IUnknown* pUnkOutput,
[in] long fAllowFormatChanges);
HRESULT _stdcall GetOutputObjectToken([out] ISpObjectToken** ppObjectToken);
HRESULT _stdcall GetOutputStream([out] ISpStreamFormat** ppStream);
HRESULT _stdcall Pause();
HRESULT _stdcall Resume();
HRESULT _stdcall SetVoice([in] ISpObjectToken* pToken);
HRESULT _stdcall GetVoice([out] ISpObjectToken** ppToken);
HRESULT _stdcall Speak(
[in] LPWSTR pwcs,
[in] unsigned long dwFlags,
[out] unsigned long* pulStreamNumber);
HRESULT _stdcall SpeakStream(
[in] IStream* pStream,
[in] unsigned long dwFlags,
[out] unsigned long* pulStreamNumber);
HRESULT _stdcall GetStatus(
[out] SPVOICESTATUS* pStatus,
[out] LPWSTR* ppszLastBookmark);
HRESULT _stdcall Skip(
[in] LPWSTR pItemType,
[in] long lNumItems,
[out] unsigned long* pulNumSkipped);
HRESULT _stdcall SetPriority([in] SPVPRIORITY ePriority);
HRESULT _stdcall GetPriority([out] SPVPRIORITY* pePriority);
HRESULT _stdcall SetAlertBoundary([in] SPEVENTENUM eBoundary);
HRESULT _stdcall GetAlertBoundary([out] SPEVENTENUM* peBoundary);
HRESULT _stdcall SetRate([in] long RateAdjust);
HRESULT _stdcall GetRate([out] long* pRateAdjust);
HRESULT _stdcall SetVolume([in] unsigned short usVolume);
HRESULT _stdcall GetVolume([out] unsigned short* pusVolume);
HRESULT _stdcall WaitUntilDone([in] unsigned long msTimeout);
HRESULT _stdcall SetSyncSpeakTimeout([in] unsigned long msTimeout);
HRESULT _stdcall GetSyncSpeakTimeout([out] unsigned long* pmsTimeout);
void* _stdcall SpeakCompleteEvent();
HRESULT _stdcall IsUISupported(
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData,
[out] long* pfSupported);
HRESULT _stdcall DisplayUI(
[in] wireHWND hWndParent,
[in] LPWSTR pszTitle,
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData);
};
typedef struct tagSPVOICESTATUS {
unsigned long ulCurrentStream;
unsigned long ulLastStreamQueued;
HRESULT hrLastResult;
unsigned long dwRunningState;
unsigned long ulInputWordPos;
unsigned long ulInputWordLen;
unsigned long ulInputSentPos;
unsigned long ulInputSentLen;
long lBookmarkId;
unsigned short PhonemeId;
SPVISEMES VisemeId;
unsigned long dwReserved1;
unsigned long dwReserved2;
} SPVOICESTATUS;
typedef enum {
SP_VISEME_0 = 0,
SP_VISEME_1 = 1,
SP_VISEME_2 = 2,
SP_VISEME_3 = 3,
SP_VISEME_4 = 4,
SP_VISEME_5 = 5,
SP_VISEME_6 = 6,
SP_VISEME_7 = 7,
SP_VISEME_8 = 8,
SP_VISEME_9 = 9,
SP_VISEME_10 = 10,
SP_VISEME_11 = 11,
SP_VISEME_12 = 12,
SP_VISEME_13 = 13,
SP_VISEME_14 = 14,
SP_VISEME_15 = 15,
SP_VISEME_16 = 16,
SP_VISEME_17 = 17,
SP_VISEME_18 = 18,
SP_VISEME_19 = 19,
SP_VISEME_20 = 20,
SP_VISEME_21 = 21
} SPVISEMES;
typedef enum {
SPVPRI_NORMAL = 0,
SPVPRI_ALERT = 1,
SPVPRI_OVER = 2
} SPVPRIORITY;
typedef enum {
SPEI_UNDEFINED = 0,
SPEI_START_INPUT_STREAM = 1,
SPEI_END_INPUT_STREAM = 2,
SPEI_VOICE_CHANGE = 3,
SPEI_TTS_BOOKMARK = 4,
SPEI_WORD_BOUNDARY = 5,
SPEI_PHONEME = 6,
SPEI_SENTENCE_BOUNDARY = 7,
SPEI_VISEME = 8,
SPEI_TTS_AUDIO_LEVEL = 9,
SPEI_TTS_PRIVATE = 15,
SPEI_MIN_TTS = 1,
SPEI_MAX_TTS = 15,
SPEI_END_SR_STREAM = 34,
SPEI_SOUND_START = 35,
SPEI_SOUND_END = 36,
SPEI_PHRASE_START = 37,
SPEI_RECOGNITION = 38,
SPEI_HYPOTHESIS = 39,
SPEI_SR_BOOKMARK = 40,
SPEI_PROPERTY_NUM_CHANGE = 41,
SPEI_PROPERTY_STRING_CHANGE = 42,
SPEI_FALSE_RECOGNITION = 43,
SPEI_INTERFERENCE = 44,
SPEI_REQUEST_UI = 45,
SPEI_RECO_STATE_CHANGE = 46,
SPEI_ADAPTATION = 47,
SPEI_START_SR_STREAM = 48,
SPEI_RECO_OTHER_CONTEXT = 49,
SPEI_SR_AUDIO_LEVEL = 50,
SPEI_SR_RETAINEDAUDIO = 51,
SPEI_SR_PRIVATE = 52,
SPEI_ACTIVE_CATEGORY_CHANGED = 53,
SPEI_RESERVED5 = 54,
SPEI_RESERVED6 = 55,
SPEI_MIN_SR = 34,
SPEI_MAX_SR = 55,
SPEI_RESERVED1 = 30,
SPEI_RESERVED2 = 33,
SPEI_RESERVED3 = 63
} SPEVENTENUM;
[
odl,
uuid(B2745EFD-42CE-48CA-81F1-A96E02538A90),
helpstring("ISpPhoneticAlphabetSelection Interface"),
restricted
]
interface ISpPhoneticAlphabetSelection : IUnknown {
HRESULT _stdcall IsAlphabetUPS([out] long* pfIsUPS);
HRESULT _stdcall SetAlphabetToUPS([in] long fForceUPS);
};
[
uuid(47206204-5ECA-11D2-960F-00C04F8EE628),
helpstring("SpSharedRecoContext Class")
]
coclass SpSharedRecoContext {
[default] interface ISpeechRecoContext;
interface ISpRecoContext;
interface ISpRecoContext2;
interface ISpPhoneticAlphabetSelection;
[default, source] dispinterface _ISpeechRecoContextEvents;
};
[
odl,
uuid(F740A62F-7C15-489E-8234-940A33D9272D),
helpstring("ISpRecoContext Interface"),
restricted
]
interface ISpRecoContext : ISpEventSource {
HRESULT _stdcall GetRecognizer([out] ISpRecognizer** ppRecognizer);
HRESULT _stdcall CreateGrammar(
[in] uint64 ullGrammarID,
[out] ISpRecoGrammar** ppGrammar);
HRESULT _stdcall GetStatus([out] SPRECOCONTEXTSTATUS* pStatus);
HRESULT _stdcall GetMaxAlternates([in] unsigned long* pcAlternates);
HRESULT _stdcall SetMaxAlternates([in] unsigned long cAlternates);
HRESULT _stdcall SetAudioOptions(
[in] SPAUDIOOPTIONS Options,
[in] GUID* pAudioFormatId,
[in] WAVEFORMATEX* pWaveFormatEx);
HRESULT _stdcall GetAudioOptions(
[in] SPAUDIOOPTIONS* pOptions,
[out] GUID* pAudioFormatId,
[out] WAVEFORMATEX** ppCoMemWFEX);
HRESULT _stdcall DeserializeResult(
[in] SPSERIALIZEDRESULT* pSerializedResult,
[out] ISpRecoResult** ppResult);
HRESULT _stdcall Bookmark(
[in] SPBOOKMARKOPTIONS Options,
[in] uint64 ullStreamPosition,
[in] LONG_PTR lparamEvent);
HRESULT _stdcall SetAdaptationData(
[in] LPWSTR pAdaptationData,
[in] unsigned long cch);
HRESULT _stdcall Pause([in] unsigned long dwReserved);
HRESULT _stdcall Resume([in] unsigned long dwReserved);
HRESULT _stdcall SetVoice(
[in] ISpVoice* pVoice,
[in] long fAllowFormatChanges);
HRESULT _stdcall GetVoice([out] ISpVoice** ppVoice);
HRESULT _stdcall SetVoicePurgeEvent([in] uint64 ullEventInterest);
HRESULT _stdcall GetVoicePurgeEvent([out] uint64* pullEventInterest);
HRESULT _stdcall SetContextState([in] SPCONTEXTSTATE eContextState);
HRESULT _stdcall GetContextState([out] SPCONTEXTSTATE* peContextState);
};
[
odl,
uuid(C2B5F241-DAA0-4507-9E16-5A1EAA2B7A5C),
helpstring("ISpRecognizer Interface"),
restricted
]
interface ISpRecognizer : ISpProperties {
HRESULT _stdcall SetRecognizer([in] ISpObjectToken* pRecognizer);
HRESULT _stdcall GetRecognizer([out] ISpObjectToken** ppRecognizer);
HRESULT _stdcall SetInput(
[in] IUnknown* pUnkInput,
[in] long fAllowFormatChanges);
HRESULT _stdcall GetInputObjectToken([out] ISpObjectToken** ppToken);
HRESULT _stdcall GetInputStream([out] ISpStreamFormat** ppStream);
HRESULT _stdcall CreateRecoContext([out] ISpRecoContext** ppNewCtxt);
HRESULT _stdcall GetRecoProfile([out] ISpObjectToken** ppToken);
HRESULT _stdcall SetRecoProfile([in] ISpObjectToken* pToken);
HRESULT _stdcall IsSharedInstance();
HRESULT _stdcall GetRecoState([out] SPRECOSTATE* pState);
HRESULT _stdcall SetRecoState([in] SPRECOSTATE NewState);
HRESULT _stdcall GetStatus([out] SPRECOGNIZERSTATUS* pStatus);
HRESULT _stdcall GetFormat(
[in] SPSTREAMFORMATTYPE WaveFormatType,
[out] GUID* pFormatId,
[out] WAVEFORMATEX** ppCoMemWFEX);
HRESULT _stdcall IsUISupported(
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData,
[out] long* pfSupported);
HRESULT _stdcall DisplayUI(
[in] wireHWND hWndParent,
[in] LPWSTR pszTitle,
[in] LPWSTR pszTypeOfUI,
[in] void* pvExtraData,
[in] unsigned long cbExtraData);
HRESULT _stdcall EmulateRecognition([in] ISpPhrase* pPhrase);
};
[
odl,
uuid(5B4FB971-B115-4DE1-AD97-E482E3BF6EE4),
helpstring("ISpProperties Interface"),
restricted
]
interface ISpProperties : IUnknown {
HRESULT _stdcall SetPropertyNum(
[in] LPWSTR pName,
[in] long lValue);
HRESULT _stdcall GetPropertyNum(
[in] LPWSTR pName,
[out] long* plValue);
HRESULT _stdcall SetPropertyString(
[in] LPWSTR pName,
[in] LPWSTR pValue);
HRESULT _stdcall GetPropertyString(
[in] LPWSTR pName,
[out] LPWSTR* ppCoMemValue);
};
typedef enum {
SPRST_INACTIVE = 0,
SPRST_ACTIVE = 1,
SPRST_ACTIVE_ALWAYS = 2,
SPRST_INACTIVE_WITH_PURGE = 3,
SPRST_NUM_STATES = 4
} SPRECOSTATE;
typedef struct tagSPRECOGNIZERSTATUS {
SPAUDIOSTATUS AudioStatus;
uint64 ullRecognitionStreamPos;
unsigned long ulStreamNumber;
unsigned long ulNumActive;
GUID ClsidEngine;
unsigned long cLangIDs;
unsigned short aLangID[20];
uint64 ullRecognitionStreamTime;
} SPRECOGNIZERSTATUS;
typedef [public]
SPWAVEFORMATTYPE SPSTREAMFORMATTYPE;
typedef enum {
SPWF_INPUT = 0,
SPWF_SRENGINE = 1
} SPWAVEFORMATTYPE;
[
odl,
uuid(1A5C0354-B621-4B5A-8791-D306ED379E53),
helpstring("ISpPhrase Interface"),
restricted
]
interface ISpPhrase : IUnknown {
HRESULT _stdcall GetPhrase([out] SPPHRASE** ppCoMemPhrase);
HRESULT _stdcall GetSerializedPhrase([out] SPSERIALIZEDPHRASE** ppCoMemPhrase);
HRESULT _stdcall GetText(
[in] unsigned long ulStart,
[in] unsigned long ulCount,
[in] long fUseTextReplacements,
[out] LPWSTR* ppszCoMemText,
[out, optional] unsigned char* pbDisplayAttributes);
HRESULT _stdcall Discard([in] unsigned long dwValueTypes);
};
typedef struct tagSPPHRASE {
unsigned long cbSize;
unsigned short LangId;
unsigned short wHomophoneGroupId;
uint64 ullGrammarID;
uint64 ftStartTime;
uint64 ullAudioStreamPosition;
unsigned long ulAudioSizeBytes;
unsigned long ulRetainedSizeBytes;
unsigned long ulAudioSizeTime;
SPPHRASERULE Rule;
SPPHRASEPROPERTY* pProperties;
SPPHRASEELEMENT* pElements;
unsigned long cReplacements;
SPPHRASEREPLACEMENT* pReplacements;
GUID SREngineID;
unsigned long ulSREnginePrivateDataSize;
unsigned char* pSREnginePrivateData;
LPWSTR pSML;
SPSEMANTICERRORINFO* pSemanticErrorInfo;
SPSEMANTICFORMAT SemanticTagFormat;
} SPPHRASE;
typedef struct tagSPPHRASERULE {
LPWSTR pszName;
unsigned long ulId;
unsigned long ulFirstElement;
unsigned long ulCountOfElements;
SPPHRASERULE* pNextSibling;
SPPHRASERULE* pFirstChild;
single SREngineConfidence;
char Confidence;
} SPPHRASERULE;
typedef struct tagSPPHRASEPROPERTY {
LPWSTR pszName;
__MIDL___MIDL_itf_sapi_0000_0020_0001 __MIDL____MIDL_itf_sapi_0000_00200001;
LPWSTR pszValue;
VARIANT vValue;
unsigned long ulFirstElement;
unsigned long ulCountOfElements;
SPPHRASEPROPERTY* pNextSibling;
SPPHRASEPROPERTY* pFirstChild;
single SREngineConfidence;
char Confidence;
} SPPHRASEPROPERTY;
typedef union tag__MIDL___MIDL_itf_sapi_0000_0020_0001 {
unsigned long ulId;
__MIDL___MIDL_itf_sapi_0000_0020_0002 __MIDL____MIDL_itf_sapi_0000_00200000;
} __MIDL___MIDL_itf_sapi_0000_0020_0001;
typedef struct tag__MIDL___MIDL_itf_sapi_0000_0020_0002 {
unsigned char bType;
unsigned char bReserved;
unsigned short usArrayIndex;
} __MIDL___MIDL_itf_sapi_0000_0020_0002;
typedef struct tagSPPHRASEELEMENT {
unsigned long ulAudioTimeOffset;
unsigned long ulAudioSizeTime;
unsigned long ulAudioStreamOffset;
unsigned long ulAudioSizeBytes;
unsigned long ulRetainedStreamOffset;
unsigned long ulRetainedSizeBytes;
LPWSTR pszDisplayText;
LPWSTR pszLexicalForm;
unsigned short* pszPronunciation;
unsigned char bDisplayAttributes;
char RequiredConfidence;
char ActualConfidence;
unsigned char reserved;
single SREngineConfidence;
} SPPHRASEELEMENT;
typedef struct tagSPPHRASEREPLACEMENT {
unsigned char bDisplayAttributes;
LPWSTR pszReplacementText;
unsigned long ulFirstElement;
unsigned long ulCountOfElements;
} SPPHRASEREPLACEMENT;
typedef struct tagSPSEMANTICERRORINFO {
unsigned long ulLineNumber;
LPWSTR pszScriptLine;
LPWSTR pszSource;
LPWSTR pszDescription;
HRESULT hrResultCode;
} SPSEMANTICERRORINFO;
typedef enum {
SPSMF_SAPI_PROPERTIES = 0,
SPSMF_SRGS_SEMANTICINTERPRETATION_MS = 1,
SPSMF_SRGS_SAPIPROPERTIES = 2,
SPSMF_UPS = 4,
SPSMF_SRGS_SEMANTICINTERPRETATION_W3C = 8
} SPSEMANTICFORMAT;
typedef struct tagSPSERIALIZEDPHRASE {
unsigned long ulSerializedSize;
} SPSERIALIZEDPHRASE;
[
odl,
uuid(2177DB29-7F45-47D0-8554-067E91C80502),
helpstring("ISpRecoGrammar Interface"),
restricted
]
interface ISpRecoGrammar : ISpGrammarBuilder {
HRESULT _stdcall GetGrammarId([out] uint64* pullGrammarId);
HRESULT _stdcall GetRecoContext([out] ISpRecoContext** ppRecoCtxt);
HRESULT _stdcall LoadCmdFromFile(
[in] LPWSTR pszFileName,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall LoadCmdFromObject(
[in] GUID* rcid,
[in] LPWSTR pszGrammarName,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall LoadCmdFromResource(
[in] void* hModule,
[in] LPWSTR pszResourceName,
[in] LPWSTR pszResourceType,
[in] unsigned short wLanguage,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall LoadCmdFromMemory(
[in] SPBINARYGRAMMAR* pGrammar,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall LoadCmdFromProprietaryGrammar(
[in] GUID* rguidParam,
[in] LPWSTR pszStringParam,
[in] void* pvDataPrarm,
[in] unsigned long cbDataSize,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall SetRuleState(
[in] LPWSTR pszName,
[in] void* pReserved,
[in] SPRULESTATE NewState);
HRESULT _stdcall SetRuleIdState(
[in] unsigned long ulRuleId,
[in] SPRULESTATE NewState);
HRESULT _stdcall LoadDictation(
[in] LPWSTR pszTopicName,
[in] SPLOADOPTIONS Options);
HRESULT _stdcall UnloadDictation();
HRESULT _stdcall SetDictationState([in] SPRULESTATE NewState);
HRESULT _stdcall SetWordSequenceData(
[in] unsigned short* pText,
[in] unsigned long cchText,
[in] SPTEXTSELECTIONINFO* pInfo);
HRESULT _stdcall SetTextSelection([in] SPTEXTSELECTIONINFO* pInfo);
HRESULT _stdcall IsPronounceable(
[in] LPWSTR pszWord,
[out] SPWORDPRONOUNCEABLE* pWordPronounceable);
HRESULT _stdcall SetGrammarState([in] SPGRAMMARSTATE eGrammarState);
HRESULT _stdcall SaveCmd(
[in] IStream* pStream,
[out, optional] LPWSTR* ppszCoMemErrorText);
HRESULT _stdcall GetGrammarState([out] SPGRAMMARSTATE* peGrammarState);
};
[
odl,
uuid(8137828F-591A-4A42-BE58-49EA7EBAAC68),
helpstring("ISpGrammarBuilder Interface"),
restricted
]
interface ISpGrammarBuilder : IUnknown {
HRESULT _stdcall ResetGrammar([in] unsigned short NewLanguage);
HRESULT _stdcall GetRule(
[in] LPWSTR pszRuleName,
[in] unsigned long dwRuleId,
[in] unsigned long dwAttributes,
[in] long fCreateIfNotExist,
[out] void** phInitialState);
HRESULT _stdcall ClearRule([in] void* hState);
HRESULT _stdcall CreateNewState(
[in] void* hState,
[out] void** phState);
HRESULT _stdcall AddWordTransition(
[in] void* hFromState,
[in] void* hToState,
[in] LPWSTR psz,
[in] LPWSTR pszSeparators,
[in] SPGRAMMARWORDTYPE eWordType,
[in] single Weight,
[in] SPPROPERTYINFO* pPropInfo);
HRESULT _stdcall AddRuleTransition(
[in] void* hFromState,
[in] void* hToState,
[in] void* hRule,
[in] single Weight,
[in] SPPROPERTYINFO* pPropInfo);
HRESULT _stdcall AddResource(
[in] void* hRuleState,
[in] LPWSTR pszResourceName,
[in] LPWSTR pszResourceValue);
HRESULT _stdcall Commit([in] unsigned long dwReserved);
};
typedef enum {
SPWT_DISPLAY = 0,
SPWT_LEXICAL = 1,
SPWT_PRONUNCIATION = 2,
SPWT_LEXICAL_NO_SPECIAL_CHARS = 3
} SPGRAMMARWORDTYPE;
typedef [public]
tagSPPROPERTYINFO SPPROPERTYINFO;
typedef struct tagtagSPPROPERTYINFO {
LPWSTR pszName;
unsigned long ulId;
LPWSTR pszValue;
VARIANT vValue;
} tagSPPROPERTYINFO;
typedef enum {
SPLO_STATIC = 0,
SPLO_DYNAMIC = 1
} SPLOADOPTIONS;
typedef struct tagSPBINARYGRAMMAR {
unsigned long ulTotalSerializedSize;
} SPBINARYGRAMMAR;
typedef enum {
SPRS_INACTIVE = 0,
SPRS_ACTIVE = 1,
SPRS_ACTIVE_WITH_AUTO_PAUSE = 3,
SPRS_ACTIVE_USER_DELIMITED = 4
} SPRULESTATE;
typedef [public]
tagSPTEXTSELECTIONINFO SPTEXTSELECTIONINFO;
typedef struct tagtagSPTEXTSELECTIONINFO {
unsigned long ulStartActiveOffset;
unsigned long cchActiveChars;
unsigned long ulStartSelection;
unsigned long cchSelection;
} tagSPTEXTSELECTIONINFO;
typedef enum {
SPWP_UNKNOWN_WORD_UNPRONOUNCEABLE = 0,
SPWP_UNKNOWN_WORD_PRONOUNCEABLE = 1,
SPWP_KNOWN_WORD_PRONOUNCEABLE = 2
} SPWORDPRONOUNCEABLE;
typedef enum {
SPGS_DISABLED = 0,
SPGS_ENABLED = 1,
SPGS_EXCLUSIVE = 3
} SPGRAMMARSTATE;
typedef struct tagSPRECOCONTEXTSTATUS {
SPINTERFERENCE eInterference;
unsigned short szRequestTypeOfUI[255];
unsigned long dwReserved1;
unsigned long dwReserved2;
} SPRECOCONTEXTSTATUS;
typedef enum {
SPINTERFERENCE_NONE = 0,
SPINTERFERENCE_NOISE = 1,
SPINTERFERENCE_NOSIGNAL = 2,
SPINTERFERENCE_TOOLOUD = 3,
SPINTERFERENCE_TOOQUIET = 4,
SPINTERFERENCE_TOOFAST = 5,
SPINTERFERENCE_TOOSLOW = 6,
SPINTERFERENCE_LATENCY_WARNING = 7,
SPINTERFERENCE_LATENCY_TRUNCATE_BEGIN = 8,
SPINTERFERENCE_LATENCY_TRUNCATE_END = 9
} SPINTERFERENCE;
typedef enum {
SPAO_NONE = 0,
SPAO_RETAIN_AUDIO = 1
} SPAUDIOOPTIONS;
typedef struct tagSPSERIALIZEDRESULT {
unsigned long ulSerializedSize;
} SPSERIALIZEDRESULT;
[
odl,
uuid(20B053BE-E235-43CD-9A2A-8D17A48B7842),
helpstring("ISpRecoResult Interface"),
restricted
]
interface ISpRecoResult : ISpPhrase {
HRESULT _stdcall GetResultTimes([out] SPRECORESULTTIMES* pTimes);
HRESULT _stdcall GetAlternates(
[in] unsigned long ulStartElement,
[in] unsigned long cElements,
[in] unsigned long ulRequestCount,
[out] ISpPhraseAlt** ppPhrases,
[out] unsigned long* pcPhrasesReturned);
HRESULT _stdcall GetAudio(
[in] unsigned long ulStartElement,
[in] unsigned long cElements,
[out] ISpStreamFormat** ppStream);
HRESULT _stdcall SpeakAudio(
[in] unsigned long ulStartElement,
[in] unsigned long cElements,
[in] unsigned long dwFlags,
[out] unsigned long* pulStreamNumber);
HRESULT _stdcall Serialize([out] SPSERIALIZEDRESULT** ppCoMemSerializedResult);
HRESULT _stdcall ScaleAudio(
[in] GUID* pAudioFormatId,
[in] WAVEFORMATEX* pWaveFormatEx);
HRESULT _stdcall GetRecoContext([out] ISpRecoContext** ppRecoContext);
};
typedef struct tagSPRECORESULTTIMES {
_FILETIME ftStreamTime;
uint64 ullLength;
unsigned long dwTickCount;
uint64 ullStart;
} SPRECORESULTTIMES;
[
odl,
uuid(8FCEBC98-4E49-4067-9C6C-D86A0E092E3D),
helpstring("ISpPhraseAlt Interface"),
restricted
]
interface ISpPhraseAlt : ISpPhrase {
HRESULT _stdcall GetAltInfo(
[out] ISpPhrase** ppParent,
[out] unsigned long* pulStartElementInParent,
[out] unsigned long* pcElementsInParent,
[out] unsigned long* pcElementsInAlt);
HRESULT _stdcall Commit();
};
typedef enum {
SPBO_NONE = 0,
SPBO_PAUSE = 1,
SPBO_AHEAD = 2,
SPBO_TIME_UNITS = 4
} SPBOOKMARKOPTIONS;
typedef enum {
SPCS_DISABLED = 0,
SPCS_ENABLED = 1
} SPCONTEXTSTATE;
[
odl,
uuid(BEAD311C-52FF-437F-9464-6B21054CA73D),
helpstring("ISpRecoContext2 Interface"),
restricted
]
interface ISpRecoContext2 : IUnknown {
HRESULT _stdcall SetGrammarOptions([in] unsigned long eGrammarOptions);
HRESULT _stdcall GetGrammarOptions([out] unsigned long* peGrammarOptions);
HRESULT _stdcall SetAdaptationData2(
[in] LPWSTR pAdaptationData,
[in] unsigned long cch,
[in] LPWSTR pTopicName,
[in] unsigned long eAdaptationSettings,
[in] SPADAPTATIONRELEVANCE eRelevance);
};
typedef enum {
SPAR_Unknown = 0,
SPAR_Low = 1,
SPAR_Medium = 2,
SPAR_High = 3
} SPADAPTATIONRELEVANCE;
[
uuid(41B89B6B-9399-11D2-9623-00C04F8EE628),
helpstring("SpInprocRecognizer Class")
]
coclass SpInprocRecognizer {
[default] interface ISpeechRecognizer;
interface ISpRecognizer;
interface ISpRecognizer2;
interface ISpRecognizer3;
interface ISpSerializeState;
};
[
odl,
uuid(8FC6D974-C81E-4098-93C5-0147F61ED4D3),
helpstring("ISpRecognizer2 Interface"),
restricted
]
interface ISpRecognizer2 : IUnknown {
HRESULT _stdcall EmulateRecognitionEx(
[in] ISpPhrase* pPhrase,
[in] unsigned long dwCompareFlags);
HRESULT _stdcall SetTrainingState(
[in] long fDoingTraining,
[in] long fAdaptFromTrainingData);
HRESULT _stdcall ResetAcousticModelAdaptation();
};
[
odl,
uuid(DF1B943C-5838-4AA2-8706-D7CD5B333499),
helpstring("ISpRecognizer3 Interface"),
restricted
]
interface ISpRecognizer3 : IUnknown {
HRESULT _stdcall GetCategory(
[in] SPCATEGORYTYPE categoryType,
[out] ISpRecoCategory** ppCategory);
HRESULT _stdcall SetActiveCategory([in] ISpRecoCategory* pCategory);
HRESULT _stdcall GetActiveCategory([out] ISpRecoCategory** ppCategory);
};
typedef enum {
SPCT_COMMAND = 0,
SPCT_DICTATION = 1,
SPCT_SLEEP = 2,
SPCT_SUB_COMMAND = 3,
SPCT_SUB_DICTATION = 4
} SPCATEGORYTYPE;
[
odl,
uuid(DA0CD0F9-14A2-4F09-8C2A-85CC48979345),
helpstring("ISpRecoCategory Interface"),
restricted
]
interface ISpRecoCategory : IUnknown {
HRESULT _stdcall GetType([out] SPCATEGORYTYPE* peCategoryType);
};
[
odl,
uuid(21B501A0-0EC7-46C9-92C3-A2BC784C54B9),
helpstring("ISpSerializeState Interface"),
restricted
]
interface ISpSerializeState : IUnknown {
HRESULT _stdcall GetSerializedState(
[out] unsigned char** ppbData,
[out] unsigned long* pulSize,
[in] unsigned long dwReserved);
HRESULT _stdcall SetSerializedState(
[in] unsigned char* pbData,
[in] unsigned long ulSize,
[in] unsigned long dwReserved);
};
[
uuid(3BEE4890-4FE9-4A37-8C1E-5E7E12791C1F),
helpstring("SpSharedRecognizer Class")
]
coclass SpSharedRecognizer {
[default] interface ISpeechRecognizer;
interface ISpRecognizer;
interface ISpRecognizer2;
interface ISpRecognizer3;
interface ISpSerializeState;
};
[
uuid(0655E396-25D0-11D3-9C26-00C04F8EF87C),
helpstring("SpLexicon Class")
]
coclass SpLexicon {
[default] interface ISpeechLexicon;
interface ISpLexicon;
interface ISpPhoneticAlphabetSelection;
};
[
odl,
uuid(DA41A7C2-5383-4DB2-916B-6C1719E3DB58),
helpstring("ISpLexicon Interface"),
restricted
]
interface ISpLexicon : IUnknown {
HRESULT _stdcall GetPronunciations(
[in] LPWSTR pszWord,
[in] unsigned short LangId,
[in] unsigned long dwFlags,
[in, out] SPWORDPRONUNCIATIONLIST* pWordPronunciationList);
HRESULT _stdcall AddPronunciation(
[in] LPWSTR pszWord,
[in] unsigned short LangId,
[in] SPPARTOFSPEECH ePartOfSpeech,
[in] LPWSTR pszPronunciation);
HRESULT _stdcall RemovePronunciation(
[in] LPWSTR pszWord,
[in] unsigned short LangId,
[in] SPPARTOFSPEECH ePartOfSpeech,
[in] LPWSTR pszPronunciation);
HRESULT _stdcall GetGeneration([out] unsigned long* pdwGeneration);
HRESULT _stdcall GetGenerationChange(
[in] unsigned long dwFlags,
[in, out] unsigned long* pdwGeneration,
[in, out] SPWORDLIST* pWordList);
HRESULT _stdcall GetWords(
[in] unsigned long dwFlags,
[in, out] unsigned long* pdwGeneration,
[in, out] unsigned long* pdwCookie,
[in, out] SPWORDLIST* pWordList);
};
typedef struct tagSPWORDPRONUNCIATIONLIST {
unsigned long ulSize;
unsigned char* pvBuffer;
SPWORDPRONUNCIATION* pFirstWordPronunciation;
} SPWORDPRONUNCIATIONLIST;
typedef struct tagSPWORDPRONUNCIATION {
SPWORDPRONUNCIATION* pNextWordPronunciation;
SPLEXICONTYPE eLexiconType;
unsigned short LangId;
unsigned short wPronunciationFlags;
SPPARTOFSPEECH ePartOfSpeech;
unsigned short szPronunciation[1];
} SPWORDPRONUNCIATION;
typedef enum {
eLEXTYPE_USER = 1,
eLEXTYPE_APP = 2,
eLEXTYPE_VENDORLEXICON = 4,
eLEXTYPE_LETTERTOSOUND = 8,
eLEXTYPE_MORPHOLOGY = 16,
eLEXTYPE_RESERVED4 = 32,
eLEXTYPE_USER_SHORTCUT = 64,
eLEXTYPE_RESERVED6 = 128,
eLEXTYPE_RESERVED7 = 256,
eLEXTYPE_RESERVED8 = 512,
eLEXTYPE_RESERVED9 = 1024,
eLEXTYPE_RESERVED10 = 2048,
eLEXTYPE_PRIVATE1 = 4096,
eLEXTYPE_PRIVATE2 = 8192,
eLEXTYPE_PRIVATE3 = 16384,
eLEXTYPE_PRIVATE4 = 32768,
eLEXTYPE_PRIVATE5 = 0x00010000,
eLEXTYPE_PRIVATE6 = 0x00020000,
eLEXTYPE_PRIVATE7 = 0x00040000,
eLEXTYPE_PRIVATE8 = 0x00080000,
eLEXTYPE_PRIVATE9 = 0x00100000,
eLEXTYPE_PRIVATE10 = 0x00200000,
eLEXTYPE_PRIVATE11 = 0x00400000,
eLEXTYPE_PRIVATE12 = 0x00800000,
eLEXTYPE_PRIVATE13 = 0x01000000,
eLEXTYPE_PRIVATE14 = 0x02000000,
eLEXTYPE_PRIVATE15 = 0x04000000,
eLEXTYPE_PRIVATE16 = 0x08000000,
eLEXTYPE_PRIVATE17 = 0x10000000,
eLEXTYPE_PRIVATE18 = 0x20000000,
eLEXTYPE_PRIVATE19 = 0x40000000,
eLEXTYPE_PRIVATE20 = 0x80000000
} SPLEXICONTYPE;
typedef enum {
SPPS_NotOverriden = 0xffffffff,
SPPS_Unknown = 0,
SPPS_Noun = 4096,
SPPS_Verb = 8192,
SPPS_Modifier = 12288,
SPPS_Function = 16384,
SPPS_Interjection = 20480,
SPPS_Noncontent = 24576,
SPPS_LMA = 28672,
SPPS_SuppressWord = 61440
} SPPARTOFSPEECH;
typedef struct tagSPWORDLIST {
unsigned long ulSize;
unsigned char* pvBuffer;
SPWORD* pFirstWord;
} SPWORDLIST;
typedef struct tagSPWORD {
SPWORD* pNextWord;
unsigned short LangId;
unsigned short wReserved;
SPWORDTYPE eWordType;
LPWSTR pszWord;
SPWORDPRONUNCIATION* pFirstWordPronunciation;
} SPWORD;
typedef enum {
eWORDTYPE_ADDED = 1,
eWORDTYPE_DELETED = 2
} SPWORDTYPE;
[
uuid(C9E37C15-DF92-4727-85D6-72E5EEB6995A),
helpstring("SpUnCompressedLexicon Class")
]
coclass SpUnCompressedLexicon {
[default] interface ISpeechLexicon;
interface ISpLexicon;
interface ISpObjectWithToken;
interface ISpPhoneticAlphabetSelection;
};
[
uuid(90903716-2F42-11D3-9C26-00C04F8EF87C),
helpstring("SpCompressedLexicon Class"),
hidden,
restricted
]
coclass SpCompressedLexicon {
[default] interface ISpLexicon;
interface ISpObjectWithToken;
};
[
uuid(0D722F1A-9FCF-4E62-96D8-6DF8F01A26AA),
helpstring("SpShortcut Class")
]
coclass SpShortcut {
[default] interface ISpShortcut;
interface ISpObjectWithToken;
};
[
odl,
uuid(3DF681E2-EA56-11D9-8BDE-F66BAD1E3F3A),
helpstring("ISpShortcut Interface"),
restricted
]
interface ISpShortcut : IUnknown {
HRESULT _stdcall AddShortcut(
[in] LPWSTR pszDisplay,
[in] unsigned short LangId,
[in] LPWSTR pszSpoken,
[in] SPSHORTCUTTYPE shType);
HRESULT _stdcall RemoveShortcut(
[in] LPWSTR pszDisplay,
[in] unsigned short LangId,
[in] LPWSTR pszSpoken,
[in] SPSHORTCUTTYPE shType);
HRESULT _stdcall GetShortcuts(
[in] unsigned short LangId,
[in, out] SPSHORTCUTPAIRLIST* pShortcutpairList);
HRESULT _stdcall GetGeneration([out] unsigned long* pdwGeneration);
HRESULT _stdcall GetWordsFromGenerationChange(
[in, out] unsigned long* pdwGeneration,
[in, out] SPWORDLIST* pWordList);
HRESULT _stdcall GetWords(
[in, out] unsigned long* pdwGeneration,
[in, out] unsigned long* pdwCookie,
[in, out] SPWORDLIST* pWordList);
HRESULT _stdcall GetShortcutsForGeneration(
[in, out] unsigned long* pdwGeneration,
[in, out] unsigned long* pdwCookie,
[in, out] SPSHORTCUTPAIRLIST* pShortcutpairList);
HRESULT _stdcall GetGenerationChange(
[in, out] unsigned long* pdwGeneration,
[in, out] SPSHORTCUTPAIRLIST* pShortcutpairList);
};
typedef enum {
SPSHT_NotOverriden = 0xffffffff,
SPSHT_Unknown = 0,
SPSHT_EMAIL = 4096,
SPSHT_OTHER = 8192,
SPPS_RESERVED1 = 12288,
SPPS_RESERVED2 = 16384,
SPPS_RESERVED3 = 20480,
SPPS_RESERVED4 = 61440
} SPSHORTCUTTYPE;
typedef struct tagSPSHORTCUTPAIRLIST {
unsigned long ulSize;
unsigned char* pvBuffer;
SPSHORTCUTPAIR* pFirstShortcutPair;
} SPSHORTCUTPAIRLIST;
typedef struct tagSPSHORTCUTPAIR {
SPSHORTCUTPAIR* pNextSHORTCUTPAIR;
unsigned short LangId;
SPSHORTCUTTYPE shType;
LPWSTR pszDisplay;
LPWSTR pszSpoken;
} SPSHORTCUTPAIR;
[
uuid(9185F743-1143-4C28-86B5-BFF14F20E5C8),
helpstring("SpPhoneConverter Class")
]
coclass SpPhoneConverter {
[default] interface ISpeechPhoneConverter;
interface ISpPhoneConverter;
interface ISpPhoneticAlphabetSelection;
};
[
odl,
uuid(8445C581-0CAC-4A38-ABFE-9B2CE2826455),
helpstring("ISpPhoneConverter Interface"),
restricted
]
interface ISpPhoneConverter : ISpObjectWithToken {
HRESULT _stdcall PhoneToId(
[in] LPWSTR pszPhone,
[out] unsigned short* pId);
HRESULT _stdcall IdToPhone(
[in] LPWSTR pId,
[out] unsigned short* pszPhone);
};
[
uuid(4F414126-DFE3-4629-99EE-797978317EAD),
helpstring("SpPhoneticAlphabetConverter Class")
]
coclass SpPhoneticAlphabetConverter {
[default] interface ISpPhoneticAlphabetConverter;
};
[
odl,
uuid(133ADCD4-19B4-4020-9FDC-842E78253B17),
helpstring("ISpPhoneticAlphabetConverter Interface"),
restricted
]
interface ISpPhoneticAlphabetConverter : IUnknown {
HRESULT _stdcall GetLangId([out] unsigned short* pLangID);
HRESULT _stdcall SetLangId([in] unsigned short LangId);
HRESULT _stdcall SAPI2UPS(
[in] unsigned short* pszSAPIId,
[out] unsigned short* pszUPSId,
[in] unsigned long cMaxLength);
HRESULT _stdcall UPS2SAPI(
[in] unsigned short* pszUPSId,
[out] unsigned short* pszSAPIId,
[in] unsigned long cMaxLength);
HRESULT _stdcall GetMaxConvertLength(
[in] unsigned long cSrcLength,
[in] long bSAPI2UPS,
[out] unsigned long* pcMaxDestLength);
};
[
uuid(455F24E9-7396-4A16-9715-7C0FDBE3EFE3),
helpstring("SpNullPhoneConverter Class"),
hidden,
restricted
]
coclass SpNullPhoneConverter {
[default] interface ISpPhoneConverter;
};
[
uuid(0F92030A-CBFD-4AB8-A164-FF5985547FF6),
helpstring("SpTextSelectionInformation Class")
]
coclass SpTextSelectionInformation {
[default] interface ISpeechTextSelectionInformation;
};
[
uuid(C23FC28D-C55F-4720-8B32-91F73C2BD5D1),
helpstring("SpPhraseInfoBuilder Class")
]
coclass SpPhraseInfoBuilder {
[default] interface ISpeechPhraseInfoBuilder;
};
[
uuid(9EF96870-E160-4792-820D-48CF0649E4EC),
helpstring("SpAudioFormat Class")
]
coclass SpAudioFormat {
[default] interface ISpeechAudioFormat;
};
[
uuid(C79A574C-63BE-44B9-801F-283F87F898BE),
helpstring("SpWaveFormatEx Class")
]
coclass SpWaveFormatEx {
[default] interface ISpeechWaveFormatEx;
};
[
uuid(73AD6842-ACE0-45E8-A4DD-8795881A2C2A),
helpstring("SpInProcRecoContext Class")
]
coclass SpInProcRecoContext {
[default] interface ISpeechRecoContext;
interface ISpRecoContext;
interface ISpRecoContext2;
interface ISpPhoneticAlphabetSelection;
[default, source] dispinterface _ISpeechRecoContextEvents;
};
[
uuid(8DBEF13F-1948-4AA8-8CF0-048EEBED95D8),
helpstring("SpCustomStream Class")
]
coclass SpCustomStream {
[default] interface ISpeechCustomStream;
interface ISpStream;
};
[
uuid(947812B3-2AE1-4644-BA86-9E90DED7EC91),
helpstring("SpFileStream Class")
]
coclass SpFileStream {
[default] interface ISpeechFileStream;
interface ISpStream;
};
[
uuid(5FB7EF7D-DFF4-468A-B6B7-2FCBD188F994),
helpstring("SpMemoryStream Class")
]
coclass SpMemoryStream {
[default] interface ISpeechMemoryStream;
interface ISpStream;
};
[
odl,
uuid(AE39362B-45A8-4074-9B9E-CCF49AA2D0B6),
helpstring("ISpXMLRecoResult Interface"),
restricted
]
interface ISpXMLRecoResult : ISpRecoResult {
HRESULT _stdcall GetXMLResult(
[out] LPWSTR* ppszCoMemXMLResult,
[in] SPXMLRESULTOPTIONS Options);
HRESULT _stdcall GetXMLErrorInfo([out] SPSEMANTICERRORINFO* pSemanticErrorInfo);
};
[
odl,
uuid(4B37BC9E-9ED6-44A3-93D3-18F022B79EC3),
helpstring("ISpRecoGrammar2 Interface"),
restricted
]
interface ISpRecoGrammar2 : IUnknown {
HRESULT _stdcall GetRules(
[out] SPRULE** ppCoMemRules,
[out] unsigned int* puNumRules);
HRESULT _stdcall LoadCmdFromFile2(
[in] LPWSTR pszFileName,
[in] SPLOADOPTIONS Options,
[in] LPWSTR pszSharingUri,
[in] LPWSTR pszBaseUri);
HRESULT _stdcall LoadCmdFromMemory2(
[in] SPBINARYGRAMMAR* pGrammar,
[in] SPLOADOPTIONS Options,
[in] LPWSTR pszSharingUri,
[in] LPWSTR pszBaseUri);
HRESULT _stdcall SetRulePriority(
[in] LPWSTR pszRuleName,
[in] unsigned long ulRuleId,
[in] int nRulePriority);
HRESULT _stdcall SetRuleWeight(
[in] LPWSTR pszRuleName,
[in] unsigned long ulRuleId,
[in] single flWeight);
HRESULT _stdcall SetDictationWeight([in] single flWeight);
HRESULT _stdcall SetGrammarLoader([in] ISpeechResourceLoader* pLoader);
HRESULT _stdcall SetSMLSecurityManager([in] IInternetSecurityManager* pSMLSecurityManager);
};
typedef struct tagSPRULE {
LPWSTR pszRuleName;
unsigned long ulRuleId;
unsigned long dwAttributes;
} SPRULE;
[
odl,
uuid(B9AC5783-FCD0-4B21-B119-B4F8DA8FD2C3),
helpstring("ISpeechResourceLoader Interface"),
dual,
oleautomation
]
interface ISpeechResourceLoader : IDispatch {
[id(0x00000001)]
HRESULT LoadResource(
[in] BSTR bstrResourceUri,
[in] VARIANT_BOOL fAlwaysReload,
[out] IUnknown** pStream,
[out] BSTR* pbstrMIMEType,
[out] VARIANT_BOOL* pfModified,
[out] BSTR* pbstrRedirectUrl);
[id(0x00000002)]
HRESULT GetLocalCopy(
[in] BSTR bstrResourceUri,
[out] BSTR* pbstrLocalPath,
[out] BSTR* pbstrMIMEType,
[out] BSTR* pbstrRedirectUrl);
[id(0x00000003)]
HRESULT ReleaseLocalCopy([in] BSTR pbstrLocalPath);
};
[
odl,
uuid(79EAC9EE-BAF9-11CE-8C82-00AA004BA90B),
helpstring("IInternetSecurityManager Interface")
]
interface IInternetSecurityManager : IUnknown {
HRESULT _stdcall SetSecuritySite([in] IInternetSecurityMgrSite* pSite);
HRESULT _stdcall GetSecuritySite([out] IInternetSecurityMgrSite** ppSite);
HRESULT _stdcall MapUrlToZone(
[in] LPWSTR pwszUrl,
[out] unsigned long* pdwZone,
[in] unsigned long dwFlags);
HRESULT _stdcall GetSecurityId(
[in] LPWSTR pwszUrl,
[out] unsigned char* pbSecurityId,
[in, out] unsigned long* pcbSecurityId,
[in] ULONG_PTR dwReserved);
HRESULT _stdcall ProcessUrlAction(
[in] LPWSTR pwszUrl,
[in] unsigned long dwAction,
[out] unsigned char* pPolicy,
[in] unsigned long cbPolicy,
[in] unsigned char* pContext,
[in] unsigned long cbContext,
[in] unsigned long dwFlags,
[in] unsigned long dwReserved);
HRESULT _stdcall QueryCustomPolicy(
[in] LPWSTR pwszUrl,
[in] GUID* guidKey,
[out] unsigned char** ppPolicy,
[out] unsigned long* pcbPolicy,
[in] unsigned char* pContext,
[in] unsigned long cbContext,
[in] unsigned long dwReserved);
HRESULT _stdcall SetZoneMapping(
[in] unsigned long dwZone,
[in] LPWSTR lpszPattern,
[in] unsigned long dwFlags);
HRESULT _stdcall GetZoneMappings(
[in] unsigned long dwZone,
[out] IEnumString** ppenumString,
[in] unsigned long dwFlags);
};
[
odl,
uuid(79EAC9ED-BAF9-11CE-8C82-00AA004BA90B),
helpstring("IInternetSecurityMgrSite Interface")
]
interface IInternetSecurityMgrSite : IUnknown {
HRESULT _stdcall GetWindow([out] wireHWND* phwnd);
HRESULT _stdcall EnableModeless([in] long fEnable);
};
typedef [public ,
custom(F914481D-9C62-4B43-9340-E9B2E6252E5F, 1)
]
uint64 ULONG_PTR;
[
odl,
uuid(00000101-0000-0000-C000-000000000046)
]
interface IEnumString : IUnknown {
HRESULT _stdcall RemoteNext(
[in] unsigned long celt,
[out] LPWSTR* rgelt,
[out] unsigned long* pceltFetched);
HRESULT _stdcall Skip([in] unsigned long celt);
HRESULT _stdcall Reset();
HRESULT _stdcall Clone([out] IEnumString** ppEnum);
};
};
主题设计 • skyfrit.com Thinkai's Blog | 保留所有权利
62 queries in 2.468 seconds |
小键盘数字 2021年12月3日 的 10:42
需求：小键盘输入数字，电脑复读，例如3.14，读三点一四。（测量输入坐标，防止输错，输入同时听语音校对）
只有一款“读霸”可以胜任，可惜win10不完全支持了。
根据楼主介绍，用tts做了一个，可以发音，但延迟太大。
代码如下：
;http://thinkai.net/p/667
;朗读CMD传入的字符串
#SingleInstance Off
#NoTrayIcon
tts := ComObjCreate(“SAPI.SpVoice”)
;tts.Voice := tts.GetVoices(“Name=VW Hui”).Item(0) ;把发音人设置成名称为VW Hui的;win10 不支持
tts.Rate := 0 ;说话速度 -10到10
tts.Volume := 100 ;音量 0到100
;txt = 一 二 三 四 五 六 七 八 九 零 点
;tts.Speak(txt) ;开始朗读
#NoEnv ; Recommended for performance and compatibility with future AutoHotkey releases.
; #Warn ; Enable warnings to assist with detecting common errors.
SendMode Input ; Recommended for new scripts due to its superior speed and reliability.
SetWorkingDir %A_ScriptDir% ; Ensures a consistent starting directory.
#Persistent
#SingleInstance Force ;定当脚本已经运行时，再次执行，会跳过对话框并自动替换旧实例, 效果类似于 Reload 命令
PAUSE::Suspend ;暂停程序发声
;F12::reload ;修改调试时使用
$Numpad0:: ;使用前缀 $ 是为避免产生关于无效循环的警告对话框 (因为此热键 “发送了它自己”). 此外, 上述动作在 释放 此按键的时候发生.
tts.Speak(“0″)
;Send, {0}
Return
$Numpad1::
SoundPlay 1.wav
Send, {1}
Return
$Numpad2::
SoundPlay 2.wav
Send, {2}
Return
$Numpad3::
SoundPlay 3.wav
Send, {3}
Return
$Numpad4::
SoundPlay 4.wav
Send, {4}
Return
$Numpad5::
SoundPlay 5.wav
Send, {5}
Return
$Numpad6::
SoundPlay 6.wav
Send, {6}
Return
$Numpad7::
SoundPlay 7.wav
Send, {7}
Return
$Numpad8::
SoundPlay 8.wav
Send, {8}
Return
$Numpad9::
SoundPlay 9.wav
Send, {9}
Return
$Numpaddot::
SoundPlay dot.wav
Send, {.}
Return
$Numpadadd::
SoundPlay add.wav
Send, {+}
Return
$Numpadsub::
SoundPlay sub.wav
Send, {-}
Return
$Numpadmult::
SoundPlay mult.wav
Send, {*}
Return
$Numpaddiv::
SoundPlay div.wav
Send, {/}
Return
$Numpadenter::
SoundPlay enter.wav
Send, {enter}
Return
复读速度跟不上，不能正常使用，好在不需要外在资源，另有其他用途。
之前写了一个用wav文件的通用，wav文件来自https://app.xunjiepdf.com/text2voice/，文字转语音，再调整发音，剪辑wav文件，代码如下：
#NoEnv ; Recommended for performance and compatibility with future AutoHotkey releases.
; #Warn ; Enable warnings to assist with detecting common errors.
SendMode Input ; Recommended for new scripts due to its superior speed and reliability.
SetWorkingDir %A_ScriptDir% ; Ensures a consistent starting directory.
#Persistent
#SingleInstance Force ;定当脚本已经运行时，再次执行，会跳过对话框并自动替换旧实例, 效果类似于 Reload 命令
;PAUSE::Suspend ;暂停程序发声
PAUSE::ExitApp ;ctrl+pause退出程序
;F12::reload ;修改调试时使用
$Numpad0:: ;使用前缀 $ 是为避免产生关于无效循环的警告对话框 (因为此热键 “发送了它自己”). 此外, 上述动作在 释放 此按键的时候发生.
SoundPlay 0.wav
Send, {0}
Return
$Numpad1::
SoundPlay 1.wav
Send, {1}
Return
$Numpad2::
SoundPlay 2.wav
Send, {2}
Return
$Numpad3::
SoundPlay 3.wav
Send, {3}
Return
$Numpad4::
SoundPlay 4.wav
Send, {4}
Return
$Numpad5::
SoundPlay 5.wav
Send, {5}
Return
$Numpad6::
SoundPlay 6.wav
Send, {6}
Return
$Numpad7::
SoundPlay 7.wav
Send, {7}
Return
$Numpad8::
SoundPlay 8.wav
Send, {8}
Return
$Numpad9::
SoundPlay 9.wav
Send, {9}
Return
$Numpaddot::
SoundPlay dot.wav
Send, {.}
Return
$Numpadadd::
SoundPlay add.wav
Send, {+}
Return
$Numpadsub::
SoundPlay sub.wav
Send, {-}
Return
$Numpadmult::
SoundPlay mult.wav
Send, {*}
Return
$Numpaddiv::
SoundPlay div.wav
Send, {/}
Return
$Numpadenter::
SoundPlay enter.wav
Send, {enter}
Return