Use a separate thread to initialize models for lazarus examples. (#1270)

So that the main thread is not blocked and the user interface is responsive.

.github/workflows/lazarus.yaml

@@ -5,6 +5,7 @@ on:
     branches:
       - master
       - lazarus
+      - fix-lazarus
     paths:
       - '.github/workflows/lazarus.yaml'
       - 'CMakeLists.txt'

lazarus-examples/generate_subtitles/generate_subtitles.lpi

@@ -160,6 +160,10 @@
         <Filename Value="my_worker.pas"/>
         <IsPartOfProject Value="True"/>
       </Unit>
+      <Unit>
+        <Filename Value="my_init.pas"/>
+        <IsPartOfProject Value="True"/>
+      </Unit>
     </Units>
   </ProjectOptions>
   <CompilerOptions>

lazarus-examples/generate_subtitles/generate_subtitles.lpr

@@ -11,7 +11,7 @@ uses
   athreads,
   {$ENDIF}
   Interfaces, // this includes the LCL widgetset
-  Forms, unit1, my_worker
+  Forms, unit1, my_worker, my_init
   { you can add units after this };
 
 {$R *.res}

lazarus-examples/generate_subtitles/my_init.pas

@@ -0,0 +1,358 @@
+unit my_init;
+
+{$mode ObjFPC}{$H+}
+
+interface
+
+uses
+  {$IFDEF UNIX}
+  cthreads,
+  cmem,
+  {$ENDIF}
+  {$IFDEF HASAMIGA}
+  athreads,
+  {$ENDIF}
+  Classes, SysUtils;
+
+type
+  TMyInitThread = class(TThread)
+  private
+    Status: AnsiString;
+    ModelDir: AnsiString;
+    procedure ShowStatus;
+
+  protected
+    procedure Execute; override;
+  public
+    Constructor Create(CreateSuspended: Boolean; ModelDirectory: AnsiString);
+  end;
+
+var
+  MyInitThread: TMyInitThread;
+
+implementation
+
+uses
+  unit1, sherpa_onnx;
+
+function CreateVad(VadFilename: AnsiString): TSherpaOnnxVoiceActivityDetector;
+var
+  Config: TSherpaOnnxVadModelConfig;
+
+  SampleRate: Integer;
+  WindowSize: Integer;
+begin
+  Initialize(Config);
+
+  SampleRate := 16000; {Please don't change it unless you know the details}
+  WindowSize := 512; {Please don't change it unless you know the details}
+
+  Config.SileroVad.Model := VadFilename;
+  Config.SileroVad.MinSpeechDuration := 0.5;
+  Config.SileroVad.MinSilenceDuration := 0.5;
+  Config.SileroVad.Threshold := 0.5;
+  Config.SileroVad.WindowSize := WindowSize;
+  Config.NumThreads:= 2;
+  Config.Debug:= True;
+  Config.Provider:= 'cpu';
+  Config.SampleRate := SampleRate;
+
+  Result := TSherpaOnnxVoiceActivityDetector.Create(Config, 30);
+end;
+
+function CreateOfflineRecognizerTransducer(
+  Tokens: AnsiString;
+  Encoder: AnsiString;
+  Decoder: AnsiString;
+  Joiner: AnsiString;
+  ModelType: AnsiString): TSherpaOnnxOfflineRecognizer;
+var
+  Config: TSherpaOnnxOfflineRecognizerConfig;
+begin
+  Initialize(Config);
+
+  Config.ModelConfig.Transducer.Encoder := Encoder;
+  Config.ModelConfig.Transducer.Decoder := Decoder;
+  Config.ModelConfig.Transducer.Joiner := Joiner;
+
+  Config.ModelConfig.ModelType := ModelType;
+  Config.ModelConfig.Tokens := Tokens;
+  Config.ModelConfig.Provider := 'cpu';
+  Config.ModelConfig.NumThreads := 2;
+  Config.ModelConfig.Debug := False;
+
+  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
+end;
+
+function CreateOfflineRecognizerTeleSpeech(
+  Tokens: AnsiString;
+  TeleSpeech: AnsiString): TSherpaOnnxOfflineRecognizer;
+var
+  Config: TSherpaOnnxOfflineRecognizerConfig;
+begin
+  Initialize(Config);
+
+  Config.ModelConfig.TeleSpeechCtc := TeleSpeech;
+
+  Config.ModelConfig.Tokens := Tokens;
+  Config.ModelConfig.Provider := 'cpu';
+  Config.ModelConfig.NumThreads := 2;
+  Config.ModelConfig.Debug := False;
+
+  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
+end;
+
+function CreateOfflineRecognizerParaformer(
+  Tokens: AnsiString;
+  Paraformer: AnsiString): TSherpaOnnxOfflineRecognizer;
+var
+  Config: TSherpaOnnxOfflineRecognizerConfig;
+begin
+  Initialize(Config);
+
+  Config.ModelConfig.Paraformer.Model := Paraformer;
+
+  Config.ModelConfig.Tokens := Tokens;
+  Config.ModelConfig.Provider := 'cpu';
+  Config.ModelConfig.NumThreads := 2;
+  Config.ModelConfig.Debug := False;
+
+  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
+end;
+
+function CreateOfflineRecognizerSenseVoice(
+  Tokens: AnsiString;
+  SenseVoice: AnsiString): TSherpaOnnxOfflineRecognizer;
+var
+  Config: TSherpaOnnxOfflineRecognizerConfig;
+begin
+  Initialize(Config);
+
+  Config.ModelConfig.SenseVoice.Model := SenseVoice;
+  Config.ModelConfig.SenseVoice.Language := 'auto';
+  Config.ModelConfig.SenseVoice.UseItn := True;
+  Config.ModelConfig.Tokens := Tokens;
+  Config.ModelConfig.Provider := 'cpu';
+  Config.ModelConfig.NumThreads := 2;
+  Config.ModelConfig.Debug := False;
+
+  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
+end;
+
+function CreateOfflineRecognizerWhisper(
+  Tokens: AnsiString;
+  WhisperEncoder: AnsiString;
+  WhisperDecoder: AnsiString): TSherpaOnnxOfflineRecognizer;
+var
+  Config: TSherpaOnnxOfflineRecognizerConfig;
+begin
+  Initialize(Config);
+
+  Config.ModelConfig.Whisper.Encoder := WhisperEncoder;
+  Config.ModelConfig.Whisper.Decoder := WhisperDecoder;
+  Config.ModelConfig.Tokens := Tokens;
+  Config.ModelConfig.Provider := 'cpu';
+  Config.ModelConfig.NumThreads := 2;
+  Config.ModelConfig.Debug := False;
+
+  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
+end;
+
+constructor TMyInitThread.Create(CreateSuspended : boolean; ModelDirectory: AnsiString);
+begin
+  inherited Create(CreateSuspended);
+  ModelDir := ModelDirectory;
+  FreeOnTerminate := True;
+end;
+
+procedure TMyInitThread.ShowStatus;
+begin
+  Form1.UpdateInitStatus(Status);
+end;
+
+procedure TMyInitThread.Execute;
+var
+  Msg: AnsiString;
+  VadFilename: AnsiString;
+  Tokens: AnsiString;
+
+  WhisperEncoder: AnsiString;
+  WhisperDecoder: AnsiString;
+
+  SenseVoice: AnsiString;
+
+  Paraformer: AnsiString;
+
+  TeleSpeech: AnsiString;
+
+  TransducerEncoder: AnsiString; // from icefall
+  TransducerDecoder: AnsiString;
+  TransducerJoiner: AnsiString;
+
+  NeMoTransducerEncoder: AnsiString;
+  NeMoTransducerDecoder: AnsiString;
+  NeMoTransducerJoiner: AnsiString;
+begin
+    VadFilename := ModelDir + 'silero_vad.onnx';
+    Tokens := ModelDir + 'tokens.txt';
+
+    {
+      Please refer to
+      https://k2-fsa.github.io/sherpa/onnx/pretrained_models/whisper/export-onnx.html#available-models
+      for a list of whisper models.
+
+      In the code, we use the normalized filename whisper-encoder.onnx, whisper-decoder.onnx, and tokens.txt
+      You need to rename the existing model files.
+
+      For instance, if you use sherpa-onnx-whisper-tiny.en, you have to do
+        mv tiny.en-tokens.txt tokens.txt
+
+        mv tiny.en-encoder.onnx whisper-encoder.onnx
+        mv tiny.en-decoder.onnx whisper-decoder.onnx
+
+        // or use the int8.onnx
+
+        mv tiny.en-encoder.int8.onnx whisper-encoder.onnx
+        mv tiny.en-decoder.int8.onnx whisper-decoder.onnx
+    }
+    WhisperEncoder := ModelDir + 'whisper-encoder.onnx';
+    WhisperDecoder := ModelDir + 'whisper-decoder.onnx';
+
+
+    {
+      Please refer to
+      https://k2-fsa.github.io/sherpa/onnx/sense-voice/pretrained.html#pre-trained-models
+      to download models for SenseVoice.
+
+      In the code, we use the normalized model name sense-voice.onnx. You have
+      to rename the downloaded model files.
+
+      For example, you need to use
+
+          mv model.onnx sense-voice.onnx
+
+          // or use the int8.onnx
+          mv model.int8.onnx sense-voice.onnx
+    }
+
+    SenseVoice := ModelDir + 'sense-voice.onnx';
+
+    {
+      Please refer to
+      https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-paraformer/index.html
+      to download paraformer models.
+
+      Note that you have to rename model.onnx or model.int8.onnx to paraformer.onnx.
+      An example is given below for the rename:
+
+        cp model.onnx paraformer.onnx
+
+        // or use int8.onnx
+        cp model.int8.onnx paraformer.onnx
+    }
+    Paraformer := ModelDir + 'paraformer.onnx';
+
+
+    {
+      please refer to
+      https://k2-fsa.github.io/sherpa/onnx/pretrained_models/telespeech/models.html
+      to download TeleSpeech models.
+
+      Note that you have to rename model files after downloading. The following
+      is an example
+
+         mv model.onnx telespeech.onnx
+
+         // or to use int8.onnx
+
+         mv model.int8.onnx telespeech.onnx
+    }
+
+    TeleSpeech := ModelDir + 'telespeech.onnx';
+
+
+    {
+      Please refer to
+      https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
+      to download an icefall offline transducer model. Note that you need to rename the
+      model files to transducer-encoder.onnx, transducer-decoder.onnx, and
+      transducer-joiner.onnx
+    }
+    TransducerEncoder := ModelDir + 'transducer-encoder.onnx';
+    TransducerDecoder := ModelDir + 'transducer-decoder.onnx';
+    TransducerJoiner := ModelDir + 'transducer-joiner.onnx';
+
+    {
+      Please visit
+      https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
+      to donwload a NeMo transducer model.
+    }
+    NeMoTransducerEncoder := ModelDir + 'nemo-transducer-encoder.onnx';
+    NeMoTransducerDecoder := ModelDir + 'nemo-transducer-decoder.onnx';
+    NeMoTransducerJoiner := ModelDir + 'nemo-transducer-joiner.onnx';
+
+    if not FileExists(VadFilename) then
+      begin
+        Status := VadFilename + ' does not exist! Please download it from' +
+          sLineBreak + 'https://github.com/k2-fsa/sherpa-onnx/tree/asr-models';
+        Synchronize(@ShowStatus);
+        Exit;
+      end;
+
+    if Form1.Vad = nil then
+      begin
+        Form1.Vad := CreateVad(VadFilename);
+      end;
+
+    if not FileExists(Tokens) then
+      begin
+        Status := Tokens + ' not found. Please download a non-streaming ASR model first!';
+        Synchronize(@ShowStatus);
+        Exit;
+      end;
+
+    if FileExists(WhisperEncoder) and FileExists(WhisperDecoder) then
+      begin
+        Form1.OfflineRecognizer := CreateOfflineRecognizerWhisper(Tokens, WhisperEncoder, WhisperDecoder);
+        Msg := 'Whisper';
+      end
+    else if FileExists(SenseVoice) then
+      begin
+        Form1.OfflineRecognizer := CreateOfflineRecognizerSenseVoice(Tokens, SenseVoice);
+        Msg := 'SenseVoice';
+      end
+    else if FileExists(Paraformer) then
+      begin
+        Form1.OfflineRecognizer := CreateOfflineRecognizerParaformer(Tokens, Paraformer);
+        Msg := 'Paraformer';
+      end
+    else if FileExists(TeleSpeech) then
+      begin
+        Form1.OfflineRecognizer := CreateOfflineRecognizerTeleSpeech(Tokens, TeleSpeech);
+        Msg := 'TeleSpeech';
+      end
+    else if FileExists(TransducerEncoder) and FileExists(TransducerDecoder) and FileExists(TransducerJoiner) then
+        begin
+          Form1.OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
+            TransducerEncoder, TransducerDecoder, TransducerJoiner, 'transducer');
+          Msg := 'Zipformer transducer';
+        end
+    else if FileExists(NeMoTransducerEncoder) and FileExists(NeMoTransducerDecoder) and FileExists(NeMoTransducerJoiner) then
+        begin
+          Form1.OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
+            NeMoTransducerEncoder, NeMoTransducerDecoder, NeMoTransducerJoiner, 'nemo_transducer');
+          Msg := 'NeMo transducer';
+        end
+    else
+      begin
+        Status := 'Please download at least one non-streaming speech recognition model first.';
+        Synchronize(@ShowStatus);
+        Exit;
+      end;
+
+    Status := 'Congratulations! The ' + Msg + ' model is initialized succesfully!';
+    Synchronize(@ShowStatus);
+end;
+
+end.
+

lazarus-examples/generate_subtitles/unit1.pas

@@ -41,6 +41,7 @@ type
       StopTime: Single;
       TotalDuration: Single);
     procedure UpdateProgress(StopTime: Single; TotalDuration: Single);
+    procedure UpdateInitStatus(Status: AnsiString);
   public
     Vad: TSherpaOnnxVoiceActivityDetector;
     OfflineRecognizer: TSherpaOnnxOfflineRecognizer;
@@ -52,7 +53,8 @@ var
 implementation
 
 uses
-  my_worker
+  my_worker,
+  my_init
   {$IFDEF DARWIN}
   ,MacOSAll
   ,CocoaAll
@@ -76,128 +78,7 @@ begin
 end;
 {$ENDIF}
 
-function CreateVad(VadFilename: AnsiString): TSherpaOnnxVoiceActivityDetector;
-var
-  Config: TSherpaOnnxVadModelConfig;
 
-  SampleRate: Integer;
-  WindowSize: Integer;
-begin
-  Initialize(Config);
-
-  SampleRate := 16000; {Please don't change it unless you know the details}
-  WindowSize := 512; {Please don't change it unless you know the details}
-
-  Config.SileroVad.Model := VadFilename;
-  Config.SileroVad.MinSpeechDuration := 0.5;
-  Config.SileroVad.MinSilenceDuration := 0.5;
-  Config.SileroVad.Threshold := 0.5;
-  Config.SileroVad.WindowSize := WindowSize;
-  Config.NumThreads:= 2;
-  Config.Debug:= True;
-  Config.Provider:= 'cpu';
-  Config.SampleRate := SampleRate;
-
-  Result := TSherpaOnnxVoiceActivityDetector.Create(Config, 30);
-end;
-
-function CreateOfflineRecognizerTransducer(
-  Tokens: AnsiString;
-  Encoder: AnsiString;
-  Decoder: AnsiString;
-  Joiner: AnsiString;
-  ModelType: AnsiString): TSherpaOnnxOfflineRecognizer;
-var
-  Config: TSherpaOnnxOfflineRecognizerConfig;
-begin
-  Initialize(Config);
-
-  Config.ModelConfig.Transducer.Encoder := Encoder;
-  Config.ModelConfig.Transducer.Decoder := Decoder;
-  Config.ModelConfig.Transducer.Joiner := Joiner;
-
-  Config.ModelConfig.ModelType := ModelType;
-  Config.ModelConfig.Tokens := Tokens;
-  Config.ModelConfig.Provider := 'cpu';
-  Config.ModelConfig.NumThreads := 2;
-  Config.ModelConfig.Debug := False;
-
-  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
-end;
-
-function CreateOfflineRecognizerTeleSpeech(
-  Tokens: AnsiString;
-  TeleSpeech: AnsiString): TSherpaOnnxOfflineRecognizer;
-var
-  Config: TSherpaOnnxOfflineRecognizerConfig;
-begin
-  Initialize(Config);
-
-  Config.ModelConfig.TeleSpeechCtc := TeleSpeech;
-
-  Config.ModelConfig.Tokens := Tokens;
-  Config.ModelConfig.Provider := 'cpu';
-  Config.ModelConfig.NumThreads := 2;
-  Config.ModelConfig.Debug := False;
-
-  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
-end;
-
-function CreateOfflineRecognizerParaformer(
-  Tokens: AnsiString;
-  Paraformer: AnsiString): TSherpaOnnxOfflineRecognizer;
-var
-  Config: TSherpaOnnxOfflineRecognizerConfig;
-begin
-  Initialize(Config);
-
-  Config.ModelConfig.Paraformer.Model := Paraformer;
-
-  Config.ModelConfig.Tokens := Tokens;
-  Config.ModelConfig.Provider := 'cpu';
-  Config.ModelConfig.NumThreads := 2;
-  Config.ModelConfig.Debug := False;
-
-  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
-end;
-
-function CreateOfflineRecognizerSenseVoice(
-  Tokens: AnsiString;
-  SenseVoice: AnsiString): TSherpaOnnxOfflineRecognizer;
-var
-  Config: TSherpaOnnxOfflineRecognizerConfig;
-begin
-  Initialize(Config);
-
-  Config.ModelConfig.SenseVoice.Model := SenseVoice;
-  Config.ModelConfig.SenseVoice.Language := 'auto';
-  Config.ModelConfig.SenseVoice.UseItn := True;
-  Config.ModelConfig.Tokens := Tokens;
-  Config.ModelConfig.Provider := 'cpu';
-  Config.ModelConfig.NumThreads := 2;
-  Config.ModelConfig.Debug := False;
-
-  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
-end;
-
-function CreateOfflineRecognizerWhisper(
-  Tokens: AnsiString;
-  WhisperEncoder: AnsiString;
-  WhisperDecoder: AnsiString): TSherpaOnnxOfflineRecognizer;
-var
-  Config: TSherpaOnnxOfflineRecognizerConfig;
-begin
-  Initialize(Config);
-
-  Config.ModelConfig.Whisper.Encoder := WhisperEncoder;
-  Config.ModelConfig.Whisper.Decoder := WhisperDecoder;
-  Config.ModelConfig.Tokens := Tokens;
-  Config.ModelConfig.Provider := 'cpu';
-  Config.ModelConfig.NumThreads := 2;
-  Config.ModelConfig.Debug := False;
-
-  Result := TSherpaOnnxOfflineRecognizer.Create(Config);
-end;
 
 {$R *.lfm}
 
@@ -256,7 +137,7 @@ end;
 
 procedure TForm1.FormClose(Sender: TObject; var CloseAction: TCloseAction);
 begin
-  if (MyWorkerThread <> nil) and not MyWorkerThread.Finished then
+  if (MyWorkerThread <> nil) and (not MyWorkerThread.Finished) then
     begin
       MyWorkerThread.Terminate;
       MyWorkerThread.WaitFor;
@@ -310,29 +191,35 @@ begin
   Form1.ResultMemo.Lines.Add(NewResult);
 end;
 
+procedure TForm1.UpdateInitStatus(Status: AnsiString);
+begin
+  if EndsStr('model is initialized succesfully!', Status) then
+    begin
+      Form1.ResultMemo.Lines.Add(Status);
+      Form1.ResultMemo.Lines.Add('Please select a 16000Hz wave file to generate subtiles');
+      Form1.ResultMemo.Lines.Add('You can download some test wave files from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models');
+      Form1.ResultMemo.Lines.Add('For instance:');
+      Form1.ResultMemo.Lines.Add('  Chinese test wave: https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/lei-jun-test.wav');
+      Form1.ResultMemo.Lines.Add('  English test wave: https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/Obama.wav');
+      FileNameEdt.Enabled := True;
+      SelectFileBtn.Enabled := True;
+
+    end
+  else
+    begin
+      ShowMessage(Status);
+      Form1.ResultMemo.Lines.Clear();
+      Form1.ResultMemo.Lines.Add('Please refer to');
+      Form1.ResultMemo.Lines.Add('https://k2-fsa.github.io/sherpa/onnx/lazarus/generate-subtitles.html#download-models');
+      Form1.ResultMemo.Lines.Add('for how to download models');
+
+      InitBtn.Enabled := True;
+    end;
+end;
+
 procedure TForm1.InitBtnClick(Sender: TObject);
 var
-  Msg: AnsiString;
   ModelDir: AnsiString;
-  VadFilename: AnsiString;
-  Tokens: AnsiString;
-
-  WhisperEncoder: AnsiString;
-  WhisperDecoder: AnsiString;
-
-  SenseVoice: AnsiString;
-
-  Paraformer: AnsiString;
-
-  TeleSpeech: AnsiString;
-
-  TransducerEncoder: AnsiString; // from icefall
-  TransducerDecoder: AnsiString;
-  TransducerJoiner: AnsiString;
-
-  NeMoTransducerEncoder: AnsiString;
-  NeMoTransducerDecoder: AnsiString;
-  NeMoTransducerJoiner: AnsiString;
 begin
   {$IFDEF DARWIN}
     ModelDir := GetResourcesPath;
@@ -340,161 +227,9 @@ begin
     ModelDir := './';
   {$ENDIF}
 
-  VadFilename := ModelDir + 'silero_vad.onnx';
+  Form1.ResultMemo.Lines.Clear();
-  Tokens := ModelDir + 'tokens.txt';
+  ResultMemo.Lines.Add('Initializing the model. Please wait...');
-
+  MyInitThread := TMyInitThread.Create(False, ModelDir);
-  {
-    Please refer to
-    https://k2-fsa.github.io/sherpa/onnx/pretrained_models/whisper/export-onnx.html#available-models
-    for a list of whisper models.
-
-    In the code, we use the normalized filename whisper-encoder.onnx, whisper-decoder.onnx, and tokens.txt
-    You need to rename the existing model files.
-
-    For instance, if you use sherpa-onnx-whisper-tiny.en, you have to do
-      mv tiny.en-tokens.txt tokens.txt
-
-      mv tiny.en-encoder.onnx whisper-encoder.onnx
-      mv tiny.en-decoder.onnx whisper-decoder.onnx
-
-      // or use the int8.onnx
-
-      mv tiny.en-encoder.int8.onnx whisper-encoder.onnx
-      mv tiny.en-decoder.int8.onnx whisper-decoder.onnx
-  }
-  WhisperEncoder := ModelDir + 'whisper-encoder.onnx';
-  WhisperDecoder := ModelDir + 'whisper-decoder.onnx';
-
-
-  {
-    Please refer to
-    https://k2-fsa.github.io/sherpa/onnx/sense-voice/pretrained.html#pre-trained-models
-    to download models for SenseVoice.
-
-    In the code, we use the normalized model name sense-voice.onnx. You have
-    to rename the downloaded model files.
-
-    For example, you need to use
-
-        mv model.onnx sense-voice.onnx
-
-        // or use the int8.onnx
-        mv model.int8.onnx sense-voice.onnx
-  }
-
-  SenseVoice := ModelDir + 'sense-voice.onnx';
-
-  {
-    Please refer to
-    https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-paraformer/index.html
-    to download paraformer models.
-
-    Note that you have to rename model.onnx or model.int8.onnx to paraformer.onnx.
-    An example is given below for the rename:
-
-      cp model.onnx paraformer.onnx
-
-      // or use int8.onnx
-      cp model.int8.onnx paraformer.onnx
-  }
-  Paraformer := ModelDir + 'paraformer.onnx';
-
-
-  {
-    please refer to
-    https://k2-fsa.github.io/sherpa/onnx/pretrained_models/telespeech/models.html
-    to download TeleSpeech models.
-
-    Note that you have to rename model files after downloading. The following
-    is an example
-
-       mv model.onnx telespeech.onnx
-
-       // or to use int8.onnx
-
-       mv model.int8.onnx telespeech.onnx
-  }
-
-  TeleSpeech := ModelDir + 'telespeech.onnx';
-
-
-  {
-    Please refer to
-    https://k2-fsa.github.io/sherpa/onnx/pretrained_models/index.html
-    to download an icefall offline transducer model. Note that you need to rename the
-    model files to transducer-encoder.onnx, transducer-decoder.onnx, and
-    transducer-joiner.onnx
-  }
-  TransducerEncoder := ModelDir + 'transducer-encoder.onnx';
-  TransducerDecoder := ModelDir + 'transducer-decoder.onnx';
-  TransducerJoiner := ModelDir + 'transducer-joiner.onnx';
-
-  {
-    Please visit
-    https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
-    to donwload a NeMo transducer model.
-  }
-  NeMoTransducerEncoder := ModelDir + 'nemo-transducer-encoder.onnx';
-  NeMoTransducerDecoder := ModelDir + 'nemo-transducer-decoder.onnx';
-  NeMoTransducerJoiner := ModelDir + 'nemo-transducer-joiner.onnx';
-
-  if not FileExists(VadFilename) then
-    begin
-      ShowMessage(VadFilename + ' does not exist! Please download it from' +
-        sLineBreak + 'https://github.com/k2-fsa/sherpa-onnx/tree/asr-models'
-      );
-      Exit;
-    end;
-
-  Self.Vad := CreateVad(VadFilename);
-
-  if not FileExists(Tokens) then
-    begin
-      ShowMessage(Tokens + ' not found. Please download a non-streaming ASR model first!');
-      Exit;
-    end;
-
-  if FileExists(WhisperEncoder) and FileExists(WhisperDecoder) then
-    begin
-      OfflineRecognizer := CreateOfflineRecognizerWhisper(Tokens, WhisperEncoder, WhisperDecoder);
-      Msg := 'Whisper';
-    end
-  else if FileExists(SenseVoice) then
-    begin
-      OfflineRecognizer := CreateOfflineRecognizerSenseVoice(Tokens, SenseVoice);
-      Msg := 'SenseVoice';
-    end
-  else if FileExists(Paraformer) then
-    begin
-      OfflineRecognizer := CreateOfflineRecognizerParaformer(Tokens, Paraformer);
-      Msg := 'Paraformer';
-    end
-  else if FileExists(TeleSpeech) then
-    begin
-      OfflineRecognizer := CreateOfflineRecognizerTeleSpeech(Tokens, TeleSpeech);
-      Msg := 'TeleSpeech';
-    end
-  else if FileExists(TransducerEncoder) and FileExists(TransducerDecoder) and FileExists(TransducerJoiner) then
-      begin
-        OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
-          TransducerEncoder, TransducerDecoder, TransducerJoiner, 'transducer');
-        Msg := 'Zipformer transducer';
-      end
-  else if FileExists(NeMoTransducerEncoder) and FileExists(NeMoTransducerDecoder) and FileExists(NeMoTransducerJoiner) then
-      begin
-        OfflineRecognizer := CreateOfflineRecognizerTransducer(Tokens,
-          NeMoTransducerEncoder, NeMoTransducerDecoder, NeMoTransducerJoiner, 'nemo_transducer');
-        Msg := 'NeMo transducer';
-      end
-  else
-    begin
-      ShowMessage('Please download at least one non-streaming speech recognition model first.');
-      Exit;
-    end;
-
-   MessageDlg('Congrat! The ' + Msg + ' model is initialized succesfully!', mtInformation, [mbOk], 0);
-   FileNameEdt.Enabled := True;
-   SelectFileBtn.Enabled := True;
   InitBtn.Enabled := False;
 end;
 

sherpa-onnx/csrc/silero-vad-model-config.cc

@@ -69,7 +69,7 @@ bool SileroVadModelConfig::Validate() const {
 std::string SileroVadModelConfig::ToString() const {
   std::ostringstream os;
 
-  os << "SilerVadModelConfig(";
+  os << "SileroVadModelConfig(";
   os << "model=\"" << model << "\", ";
   os << "threshold=" << threshold << ", ";
   os << "min_silence_duration=" << min_silence_duration << ", ";

sherpa-onnx/pascal-api/sherpa_onnx.pas

@@ -98,6 +98,7 @@ type
     destructor Destroy; override;
     procedure AcceptWaveform(Samples: array of Single; SampleRate: Integer);
     procedure InputFinished;
+    property GetHandle: Pointer Read Handle;
   end;
 
   TSherpaOnnxOnlineRecognizer = class
@@ -116,6 +117,7 @@ type
     function IsEndpoint(Stream: TSherpaOnnxOnlineStream): Boolean;
     function GetResult(Stream: TSherpaOnnxOnlineStream): TSherpaOnnxOnlineRecognizerResult;
     property Config: TSherpaOnnxOnlineRecognizerConfig Read _Config;
+    property GetHandle: Pointer Read Handle;
   end;
 
   TSherpaOnnxOfflineTransducerModelConfig = record
@@ -213,6 +215,7 @@ type
     constructor Create(P: Pointer);
     destructor Destroy; override;
     procedure AcceptWaveform(Samples: array of Single; SampleRate: Integer);
+    property GetHandle: Pointer Read Handle;
   end;
 
   TSherpaOnnxOfflineRecognizer = class
@@ -226,6 +229,7 @@ type
     procedure Decode(Stream: TSherpaOnnxOfflineStream);
     function GetResult(Stream: TSherpaOnnxOfflineStream): TSherpaOnnxOfflineRecognizerResult;
     property Config: TSherpaOnnxOfflineRecognizerConfig Read _Config;
+    property GetHandle: Pointer Read Handle;
   end;
 
   TSherpaOnnxSileroVadModelConfig = record
@@ -262,6 +266,7 @@ type
     procedure Reset;
     function Size: Integer;
     function Head: Integer;
+    property GetHandle: Pointer Read Handle;
   end;
 
   TSherpaOnnxSpeechSegment = record
@@ -286,6 +291,7 @@ type
     procedure Reset;
     procedure Flush;
     property Config: TSherpaOnnxVadModelConfig Read _Config;
+    property GetHandle: Pointer Read Handle;
   end;
 
   { It supports reading a single channel wave with 16-bit encoded samples.