Inheritance diagram for ailiaSpeech.AiliaSpeechModel:

Collaboration diagram for ailiaSpeech.AiliaSpeechModel:

Public Member Functions
bool	IsError ()
	Check is error occured. More...

string	GetErrorDetail ()
	Get error detail. More...

int	GetEnvironmentId (bool is_gpu)
	Get the environmen id. More...

string	GetEnvironmentName ()
	Get the environmen name. More...

bool	Open (string encoder_path, string decoder_path, int env_id, int memory_mode, int model_type, int task, int flag, string language)
	Create a instance. More...

bool	OpenVad (string vad_path, int vad_type)
	Open VAD file. More...

bool	OpenDictionary (string dictionary_path, int dictionary_type)
	Open dictionary file. More...

bool	OpenPostProcess (string encoder_path, string decoder_path, string source_path, string target_path, string prefix, int type)
	Open PostProcess file. More...

bool	OpenDiarization (string segmentation_path, string embedding_path, int type)
	Open AI model for speaker diarization. More...

virtual void	Close ()
	Destroys instance. More...

virtual void	Dispose ()
	Release resources. More...

bool	SetPrompt (string prompt)
	Set prompt. More...

bool	SetConstraint (string constraint, int constraint_type)
	Set constraint. More...

bool	Transcribe (float[] waveData, uint frequency, uint channels, bool tail)
	Perform speech recognition. More...

bool	IsProcessing ()
	Check is processing sub thread. More...

bool	IsTranscribing ()
	Check is running Speech2Text. More...

bool	IsCompleted ()
	Gets whether all audio processing is complete. More...

List< string >	GetResults ()
	Get results and clear of Speech2Text. More...

List< AiliaSpeechText >	GetStructuredResults ()
	Get structured results and clear of Speech2Text. More...

string	GetIntermediateText ()
	Get the intermediate result of Speech2Text. More...

bool	ResetTranscribeState ()
	Initialize the Speech2Text state. More...

Static Public Member Functions
static int	IntermediateCallback (IntPtr handle, IntPtr text)

Protected Member Functions
virtual void	Dispose (bool disposing)

Member Function Documentation

◆ Close()

virtual void ailiaSpeech.AiliaSpeechModel.Close ( )

inlinevirtual

Destroys instance.

Destroys and initializes the instance.

     {
         DestroyThread();
         DestroyInterrupt();
         if (net != IntPtr.Zero)
         {
             AiliaSpeech.ailiaSpeechDestroy(net);
             net = IntPtr.Zero;
         }
     }

◆ Dispose() [1/2]

virtual void ailiaSpeech.AiliaSpeechModel.Dispose ( )

inlinevirtual

Release resources.

     {
         Dispose(true);
     }

◆ Dispose() [2/2]

virtual void ailiaSpeech.AiliaSpeechModel.Dispose ( bool disposing )

inlineprotectedvirtual

     {
         if (disposing){
             // release managed resource
         }
         Close(); // release unmanaged resource
     }

◆ GetEnvironmentId()

int ailiaSpeech.AiliaSpeechModel.GetEnvironmentId ( bool is_gpu )

inline

Get the environmen id.

Parameters

is_gpu Whether to use GPU

Returns: env_id

                                             {
         int env_id = Ailia.AILIA_ENVIRONMENT_ID_AUTO;
         if (is_gpu) { // GPU
             int count = 0;
             Ailia.ailiaGetEnvironmentCount(ref count);
             for (int i = 0; i < count; i++){
                 IntPtr env_ptr = IntPtr.Zero;
                 Ailia.ailiaGetEnvironment(ref env_ptr, (uint)i, Ailia.AILIA_ENVIRONMENT_VERSION);
                 Ailia.AILIAEnvironment env = (Ailia.AILIAEnvironment)Marshal.PtrToStructure(env_ptr, typeof(Ailia.AILIAEnvironment));
  
                 if (env.backend == Ailia.AILIA_ENVIRONMENT_BACKEND_MPS || env.backend == Ailia.AILIA_ENVIRONMENT_BACKEND_CUDA || env.backend == Ailia.AILIA_ENVIRONMENT_BACKEND_VULKAN){
                     env_id = env.id;
                     env_name = Marshal.PtrToStringAnsi(env.name);
                 }
             }
         } else {
             env_name = "cpu";
         }
         return env_id;
     }

◆ GetEnvironmentName()

string ailiaSpeech.AiliaSpeechModel.GetEnvironmentName ( )

inline

Get the environmen name.

Returns: env_name

                                       {
         return env_name;
     }

◆ GetErrorDetail()

string ailiaSpeech.AiliaSpeechModel.GetErrorDetail ( )

inline

Get error detail.

Returns: The error detail string.

                                   {
         return m_error_detail;
     }

◆ GetIntermediateText()

string ailiaSpeech.AiliaSpeechModel.GetIntermediateText ( )

inline

Get the intermediate result of Speech2Text.

Returns: Transcribe results.

                                        {
         lock (m_intermediate_lock_async){
             return m_intermediate_text;
         }
     }

◆ GetResults()

List<string> ailiaSpeech.AiliaSpeechModel.GetResults ( )

inline

Get results and clear of Speech2Text.

Returns: Transcribe results.

                                     {
         lock (m_lock_async)
         {
             List<string> results = new List<string>(m_results);
             m_results.Clear();
             return results;
         }
     }

◆ GetStructuredResults()

List<AiliaSpeechText> ailiaSpeech.AiliaSpeechModel.GetStructuredResults ( )

inline

Get structured results and clear of Speech2Text.

Returns: Transcribe results.

         {
             lock (m_lock_async)
             {
                 List<AiliaSpeechText> results_struct = new List<AiliaSpeechText>(m_results_struct);
                 m_results_struct.Clear();
                 return results_struct;
             }
         }

◆ IntermediateCallback()

static int ailiaSpeech.AiliaSpeechModel.IntermediateCallback	(	IntPtr	handle,
		IntPtr	text
	)

inlinestatic

                                                                       {
         lock (m_intermediate_lock_async){
             try{
                  string decoded_text = Marshal.PtrToStringAnsi(text);
                  m_intermediate_text = decoded_text;
             }catch(Exception e){
             }
         }
         if (Marshal.ReadInt32(handle) != 0){
             return -1; // 中断
         }
         return 0;
     }

◆ IsCompleted()

bool ailiaSpeech.AiliaSpeechModel.IsCompleted ( )

inline

Gets whether all audio processing is complete.

Returns: If Speech2Text is complete, it returns true , or false otherwise.

                              {
         lock (m_lock_async)
         {
             return m_complete;
         }
     }

◆ IsError()

bool ailiaSpeech.AiliaSpeechModel.IsError ( )

inline

Check is error occured.

Returns: If error is occured, it returns true , or false otherwise.

                          {
         return m_error;
     }

◆ IsProcessing()

bool ailiaSpeech.AiliaSpeechModel.IsProcessing ( )

inline

Check is processing sub thread.

Returns: If sub thread is processing, it returns true , or false otherwise.

                               {
         lock (m_lock_async)
         {
             return m_processing;
         }
     }

◆ IsTranscribing()

bool ailiaSpeech.AiliaSpeechModel.IsTranscribing ( )

inline

Check is running Speech2Text.

Returns: If Speech2Text is running, it returns true , or false otherwise.

                                 {
         lock (m_lock_async)
         {
             return m_decoding;
         }
     }

◆ Open()

bool ailiaSpeech.AiliaSpeechModel.Open	(	string	encoder_path,
		string	decoder_path,
		int	env_id,
		int	memory_mode,
		int	model_type,
		int	task,
		int	flag,
		string	language
	)

inline

Create a instance.

Parameters

encoder_path	Encoder onnx file path
decoder_path	Decoder onnx file path
env_id	Runtime environment (Ailia.AILIA_ENVIRONMENT_ID_AUTO for automatic selection)
memory_mode	Memory mode (Ailia.AILIA_MEMORY_REDUCE_CONSTANT \| Ailia.AILIA_MEMORY_REDUCE_CONSTANT_WITH_INPUT_INITIALIZER \| Ailia.AILIA_MEMORY_REUSE_INTERSTAGE etc)
model_type	Model type (AiliaSpeech.AILIA_SPEECH_MODEL_TYPE_*)
task	Task (AiliaSpeech.AILIA_SPEECH_TASK_*)
flag	OR of flags (AiliaSpeech.AILIA_SPEECH_FLAG_*)
language	Language (ja or en or etc. auto is automatic selection)

Returns: If this function is successful, it returns true , or false otherwise.

                                                                                                                                                 {
         AiliaLicense.CheckAndDownloadLicense();
  
         if (net != null){
             Close();
         }
  
         AiliaSpeech.AILIASpeechApiCallback callback = AiliaSpeech.GetCallback();
  
         int status = AiliaSpeech.ailiaSpeechCreate(ref net, env_id, Ailia.AILIA_MULTITHREAD_AUTO, memory_mode, task, flag, callback, AiliaSpeech.AILIA_SPEECH_API_CALLBACK_VERSION);
         Check(status, "ailiaSpeechCreate");
         if (status != 0){
             return false;
         }
  
         status = AiliaSpeech.ailiaSpeechOpenModelFile(net, encoder_path, decoder_path, model_type);
         Check(status, "ailiaSpeechOpenModelFile");
         if (status != 0){
             return false;
         }
  
         if (language != "auto"){
             status = AiliaSpeech.ailiaSpeechSetLanguage(net, language);
             Check(status, "ailiaSpeechSetLanguage");
             if (status != 0){
                 return false;
             }
         }
  
         status = AiliaSpeech.ailiaSpeechSetSilentThreshold(net, THRESHOLD_VOLUME, SPEECH_SEC, NO_SPEECH_SEC);
         Check(status, "ailiaSpeechSetSilentThreshold");
         if (status != 0){
             return false;
         }
  
         CreateInterrupt();
  
         status = AiliaSpeech.ailiaSpeechSetIntermediateCallback(net, IntermediateCallback, m_interrupt_ptr);
         Check(status, "ailiaSpeechSetIntermediateCallback");
         if (status != 0){
             return false;
         }
  
         CreateThread();
  
         m_error = false;
         m_error_detail = "";
  
         if ((flag & AiliaSpeech.AILIA_SPEECH_FLAG_LIVE) != 0){
             live_mode = true;
         }else{
             live_mode = false;
         }
  
         return true;
     }

◆ OpenDiarization()

bool ailiaSpeech.AiliaSpeechModel.OpenDiarization	(	string	segmentation_path,
		string	embedding_path,
		int	type
	)

inline

Open AI model for speaker diarization.

Parameters

segmentation_path	The path name to the onnx file
embedding_path	The path name to the onnx file
type	AILIA_SPEECH_DIARIZATION_TYPE_PYANNOTE_AUDIO

Returns: If this function is successful, it returns true , or false otherwise.

                                                                                           {
         if (net == null){
             return false;
         }
         int status;
         status = AiliaSpeech.ailiaSpeechOpenDiarizationFile(net, segmentation_path, embedding_path, type);
         Check(status, "ailiaSpeechOpenDializationFile");
         if (status != 0){
             return false;
         }
         return true;
     }

◆ OpenDictionary()

bool ailiaSpeech.AiliaSpeechModel.OpenDictionary	(	string	dictionary_path,
		int	dictionary_type
	)

inline

Open dictionary file.

Parameters

dictionary_path	dictionary file path
dictionary_type	dictionary type (AiliaSpeech.AILIA_SPEECH_DICTIONARY_TYPE_*)

Returns: If this function is successful, it returns true , or false otherwise.

                                                                            {
         if (net == null){
             return false;
         }
         int status = AiliaSpeech.ailiaSpeechOpenDictionaryFile(net, dictionary_path, dictionary_type);
         Check(status, "ailiaSpeechOpenDictionaryFile");
         if (status != 0){
             return false;
         }
         return true;
     }

◆ OpenPostProcess()

bool ailiaSpeech.AiliaSpeechModel.OpenPostProcess	(	string	encoder_path,
		string	decoder_path,
		string	source_path,
		string	target_path,
		string	prefix,
		int	type
	)

inline

Open PostProcess file.

Parameters

encoder_path	The path name to the onnx file
decoder_path	The path name to the onnx file
source_path	The path name to the tokenizer model file
target_path	The path name to the tokenizer model file
prefix	The prefix of T5 (UTF8), null for FuguMT
post_process_type	AILIA_SPEECH_POST_PROCESS_TYPE_*

Returns: If this function is successful, it returns true , or false otherwise.

                                                                                                                                           {
         if (net == null){
             return false;
         }
         int status;
         if (prefix == null){
             status = AiliaSpeech.ailiaSpeechOpenPostProcessFile(net, encoder_path, decoder_path, source_path, target_path, IntPtr.Zero, type);
         }else{
             byte[] text = System.Text.Encoding.UTF8.GetBytes(prefix+"\u0000");
             GCHandle handle = GCHandle.Alloc(text, GCHandleType.Pinned);
             IntPtr prefix_ptr = handle.AddrOfPinnedObject();
             status = AiliaSpeech.ailiaSpeechOpenPostProcessFile(net, encoder_path, decoder_path, source_path, target_path, prefix_ptr, type);
             handle.Free();
         }
         Check(status, "ailiaSpeechOpenPostProcessFile");
         if (status != 0){
             return false;
         }
         post_process_mode = true;
         return true;
     }

◆ OpenVad()

bool ailiaSpeech.AiliaSpeechModel.OpenVad	(	string	vad_path,
		int	vad_type
	)

inline

Open VAD file.

Parameters

vad_path	VAD onnx file path
vad_type	VAD type (AiliaSpeech.AILIA_SPEECH_VAD_TYPE_*)

Returns: If this function is successful, it returns true , or false otherwise.

                                                       {
         if (net == null){
             return false;
         }
         int status = AiliaSpeech.ailiaSpeechOpenVadFile(net, vad_path, vad_type);
         Check(status, "ailiaSpeechOpenVadFile");
         if (status != 0){
             return false;
         }
         status = AiliaSpeech.ailiaSpeechSetSilentThreshold(net, THRESHOLD_VAD, SPEECH_SEC, NO_SPEECH_SEC);
         Check(status, "ailiaSpeechSetSilentThreshold");
         if (status != 0){
             return false;
         }
         return true;
     }

◆ ResetTranscribeState()

bool ailiaSpeech.AiliaSpeechModel.ResetTranscribeState ( )

inline

Initialize the Speech2Text state.

Returns: If this function is successful, it returns true , or false otherwise.

                                       {
         int status = AiliaSpeech.ailiaSpeechResetTranscribeState(net);
         lock (m_lock_async)
         {
             m_complete = false;
         }
         if (status == 0){
             return true;
         }
         return false;
     }

◆ SetConstraint()

bool ailiaSpeech.AiliaSpeechModel.SetConstraint	(	string	constraint,
		int	constraint_type
	)

inline

Set constraint.

Parameters

constraint	The text of constraint (UTF8)
constraint_type	AILIA_SPEECH_CONSTRAINT_*

Returns: If this function is successful, it returns true , or false otherwise.

                                                                      {
         byte[] text = System.Text.Encoding.UTF8.GetBytes(constraint+"\u0000");
         GCHandle handle = GCHandle.Alloc(text, GCHandleType.Pinned);
         IntPtr input = handle.AddrOfPinnedObject();
         int status = AiliaSpeech.ailiaSpeechSetConstraint(net, input, constraint_type);
         handle.Free();
         if (status != 0){
             return false;
         }
         return true;
     }

◆ SetPrompt()

bool ailiaSpeech.AiliaSpeechModel.SetPrompt ( string prompt )

inline

Set prompt.

Parameters

prompt The text of prompt (UTF8)

Returns: If this function is successful, it returns true , or false otherwise.

                                         {
         byte[] text = System.Text.Encoding.UTF8.GetBytes(prompt+"\u0000");
         GCHandle handle = GCHandle.Alloc(text, GCHandleType.Pinned);
         IntPtr input = handle.AddrOfPinnedObject();
         int status = AiliaSpeech.ailiaSpeechSetPrompt(net, input);
         handle.Free();
         if (status != 0){
             return false;
         }
         return true;
     }

◆ Transcribe()

bool ailiaSpeech.AiliaSpeechModel.Transcribe	(	float[]	waveData,
		uint	frequency,
		uint	channels,
		bool	tail
	)

inline

Perform speech recognition.

Parameters

waveQueue	Input PCM
frequency	Frequency of PCM
channels	Number of channels of PCM
tail	Is last input

Returns: If this function is successful, it returns true , or false otherwise. @detail Run speech recognition. Non -blocked API. When the execution is completed, the IsTranscribed API returns True. Execution results can be obtained with the GetResults API. The result of execution can be obtained with the GetIntermediateText API.

     {
         if (waveData.Length == 0){
             return false;
         }
         lock (m_lock_async)
         {
             threadChannels = channels;
             threadFrequency = frequency;
             threadWaveQueue.Add(waveData);
             threadComplete = tail;
             m_processing = true;
             m_auto_event.Set();
         }
         return true;
     }

The documentation for this class was generated from the following file:

AiliaSpeechModel.cs

Public Member Functions

Static Public Member Functions

Protected Member Functions

Member Function Documentation

◆ Close()

◆ Dispose() [1/2]

◆ Dispose() [2/2]

◆ GetEnvironmentId()

◆ GetEnvironmentName()

◆ GetErrorDetail()

◆ GetIntermediateText()

◆ GetResults()

◆ GetStructuredResults()

◆ IntermediateCallback()

◆ IsCompleted()

◆ IsError()

◆ IsProcessing()

◆ IsTranscribing()

◆ Open()

◆ OpenDiarization()

◆ OpenDictionary()

◆ OpenPostProcess()

◆ OpenVad()

◆ ResetTranscribeState()

◆ SetConstraint()

◆ SetPrompt()

◆ Transcribe()