Various refinements

* Filter out transcriptions like " (music)"
* Whisper mic choice auto-populates with queried values
  * No more manually lining up numbers!
* Persist whisper mic in config
* Remove setup and dump mics button from Whisper page
  * Redesign makes these unnecessary

5 files changed, 74 insertions, 51 deletions

diff --git a/GUI/GUI/GUI/Config.cpp b/GUI/GUI/GUI/Config.cpp
index 1287619..50f0dca 100644
--- a/GUI/GUI/GUI/Config.cpp
+++ b/GUI/GUI/GUI/Config.cpp
@@ -89,7 +89,8 @@ AppConfig::AppConfig()
 	menu_path(),

 	clear_osc(false),

 

-	whisper_model("ggml-base.en.bin")

+	whisper_model("ggml-base.en.bin"),

+	whisper_mic(0)

 {}

 

 bool AppConfig::Serialize(const std::filesystem::path& path) {

@@ -118,6 +119,7 @@ bool AppConfig::Serialize(const std::filesystem::path& path) {
 	root["clear_osc"] << clear_osc;

 

 	root["whisper_model"] << whisper_model;

+	root["whisper_mic"] << whisper_mic;

 

   return Config::Serialize(path, &t);

 }

@@ -159,6 +161,7 @@ bool AppConfig::Deserialize(const std::filesystem::path& path) {
 	root.get_if("clear_osc", &c.clear_osc);

 

 	root.get_if("whisper_model", &c.whisper_model);

+	root.get_if("whisper_mic", &c.whisper_mic);

 

 	*this = std::move(c);

 	return true;

diff --git a/GUI/GUI/GUI/Config.h b/GUI/GUI/GUI/Config.h
index 07a4d8c..b27dcc5 100644
--- a/GUI/GUI/GUI/Config.h
+++ b/GUI/GUI/GUI/Config.h
@@ -63,5 +63,6 @@ public:
 

 	// WhisperCPP-specific settings.

 	std::string whisper_model;

+	int whisper_mic;

 };

 

diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index e0713c1..5332490 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -72,7 +72,7 @@ namespace {
 		ID_WHISPER_OUT,

 		ID_WHISPER_CONFIG_PANEL,

 		ID_WHISPER_SETUP_BUTTON,

-		ID_WHISPER_DUMP_MICS_BUTTON,

+		//ID_WHISPER_DUMP_MICS_BUTTON,

 		ID_WHISPER_CONFIG_PANEL_PAIRS,

 		ID_WHISPER_MIC,

 		ID_WHISPER_LANG,

@@ -108,6 +108,21 @@ namespace {
     const size_t kNumMicChoices = sizeof(kMicChoices) / sizeof(kMicChoices[0]);

     constexpr int kMicDefault = 0;  // index

 

+    wxString kWhisperMicChoices[] = {

+        "0",

+        "1",

+        "2",

+        "3",

+        "4",

+        "5",

+        "6",

+        "7",

+        "8",

+        "9",

+    };

+    const size_t kNumWhisperMicChoices = sizeof(kWhisperMicChoices) / sizeof(kWhisperMicChoices[0]);

+    constexpr int kWhisperMicDefault = 0;  // index

+

     // lifted from whisper/tokenizer.py

 	const wxString kLangChoices[] = {

 	    "english",

@@ -322,25 +337,25 @@ Frame::Frame()
         auto* navbar = new wxPanel(main_panel, ID_NAVBAR);

         {

 			auto* navbar_button_transcribe = new wxButton(navbar,

-                ID_NAVBAR_BUTTON_TRANSCRIBE, "Transcription");

+                ID_NAVBAR_BUTTON_TRANSCRIBE, "Transcription (PY)");

 			auto* navbar_button_unity = new wxButton(navbar,

                 ID_NAVBAR_BUTTON_UNITY, "Unity");

 			auto* navbar_button_debug = new wxButton(navbar,

                 ID_NAVBAR_BUTTON_DEBUG, "Debug");

 			auto* navbar_button_whisper = new wxButton(navbar,

-                ID_NAVBAR_BUTTON_WHISPER, "WhisperCPP");

+                ID_NAVBAR_BUTTON_WHISPER, "Transcription (CPP)");

 

 			auto* sizer = new wxBoxSizer(wxVERTICAL);

 			navbar->SetSizer(sizer);

 

 			sizer->Add(navbar_button_transcribe, /*proportion=*/0,

                 /*flags=*/wxEXPAND);

+			sizer->Add(navbar_button_whisper, /*proportion=*/0,

+                /*flags=*/wxEXPAND);

 			sizer->Add(navbar_button_unity, /*proportion=*/0,

                 /*flags=*/wxEXPAND);

 			sizer->Add(navbar_button_debug, /*proportion=*/0,

                 /*flags=*/wxEXPAND);

-			sizer->Add(navbar_button_whisper, /*proportion=*/0,

-                /*flags=*/wxEXPAND);

         }

 

         auto* transcribe_panel = new wxPanel(main_panel, ID_PY_PANEL);

@@ -843,27 +858,12 @@ Frame::Frame()
             auto* whisper_config_panel = new wxPanel(whisper_panel,

                 ID_WHISPER_CONFIG_PANEL);

             {

-                auto* whisper_setup_button = new wxButton(whisper_config_panel,

-                    ID_WHISPER_SETUP_BUTTON, "Set up Python virtual environment");

-                whisper_setup_button->SetToolTip(

-                    "TaSTT uses the Python programming language to provide both "

-                    "transcription services and to interface with Unity. "

-                    "It installs its dependencies into an isolated folder "

-                    "called a 'virtual environment'. Click this button to "

-                    "install those dependencies. This only has to be done "

-                    "once when you install a new version of TaSTT.");

-                auto* whisper_dump_mics_button = new wxButton(whisper_config_panel,

-                    ID_WHISPER_DUMP_MICS_BUTTON, "List input devices");

-                whisper_dump_mics_button->SetToolTip(

-                    "List the microphones (and input devices) attached to "

-                    "your computer. To use a microphone, enter the number "

-                    "to its left in the 'Microphone' dropdown.");

                 auto* whisper_config_panel_pairs = new wxPanel(whisper_config_panel,

                     ID_WHISPER_CONFIG_PANEL_PAIRS);

                 {

                     auto* whisper_mic = new wxChoice(whisper_config_panel_pairs,

                         ID_WHISPER_MIC, wxDefaultPosition,

-                        wxDefaultSize, kNumMicChoices, kMicChoices);

+                        wxDefaultSize, kNumWhisperMicChoices, kWhisperMicChoices);

                     whisper_mic->SetToolTip(

                         "Select which microphone to listen to when "

                         "transcribing. To get list microphones and get their "

@@ -1037,10 +1037,6 @@ Frame::Frame()
 

                 auto* sizer = new wxBoxSizer(wxVERTICAL);

                 whisper_config_panel->SetSizer(sizer);

-                sizer->Add(whisper_setup_button, /*proportion=*/0,

-                    /*flags=*/wxEXPAND);

-                sizer->Add(whisper_dump_mics_button, /*proportion=*/0,

-                    /*flags=*/wxEXPAND);

                 sizer->Add(whisper_config_panel_pairs, /*proportion=*/0,

                     /*flags=*/wxEXPAND);

                 sizer->Add(whisper_enable_local_beep, /*proportion=*/0,

@@ -1060,8 +1056,8 @@ Frame::Frame()
             sizer->Add(whisper_config_panel, /*proportion=*/0, /*flags=*/wxEXPAND);

             sizer->Add(whisper_out, /*proportion=*/1, /*flags=*/wxEXPAND);

         }

-        whisper_panel_->Hide();

         whisper_ = std::make_unique<WhisperCPP>(whisper_out_);

+        whisper_panel_->Hide();

 

         auto* debug_panel = new wxPanel(main_panel, ID_DEBUG_PANEL);

         debug_panel_ = debug_panel;

@@ -1180,7 +1176,7 @@ Frame::Frame()
     Bind(wxEVT_TIMER,  &Frame::OnAppDrain, this, ID_PY_APP_DRAIN);

 	Bind(wxEVT_BUTTON, &Frame::OnSetupPython, this, ID_PY_SETUP_BUTTON);

 	Bind(wxEVT_BUTTON, &Frame::OnDumpMics, this, ID_PY_DUMP_MICS_BUTTON);

-	Bind(wxEVT_BUTTON, &Frame::OnWhisperDumpMics, this, ID_WHISPER_DUMP_MICS_BUTTON);

+	//Bind(wxEVT_BUTTON, &Frame::OnWhisperDumpMics, this, ID_WHISPER_DUMP_MICS_BUTTON);

 	Bind(wxEVT_BUTTON, &Frame::OnGenerateFX, this,

         ID_UNITY_BUTTON_GEN_ANIMATOR);

 	Bind(wxEVT_BUTTON, &Frame::OnListPip, this, ID_DEBUG_BUTTON_LIST_PIP);

@@ -1205,6 +1201,7 @@ Frame::Frame()
 

     // Initialize input fields using AppConfig.

     ApplyConfigToInputFields();

+    PopulateDynamicInputFields();

 

     Resize();

 	OnUnityParamChangeImpl();

@@ -1258,7 +1255,8 @@ void Frame::ApplyConfigToInputFields()
 

     // Whisper panel

     auto* whisper_mic = static_cast<wxChoice*>(FindWindowById(ID_WHISPER_MIC));

-	whisper_mic->SetSelection(mic_idx);

+    int whisper_mic_idx = app_c_.whisper_mic;

+	whisper_mic->SetSelection(whisper_mic_idx);

 

     auto* whisper_lang = static_cast<wxChoice*>(FindWindowById(ID_WHISPER_LANG));

 	whisper_lang->SetSelection(lang_idx);

@@ -1266,7 +1264,7 @@ void Frame::ApplyConfigToInputFields()
     auto* whisper_model = static_cast<wxChoice*>(FindWindowById(ID_WHISPER_MODEL));

 	int whisper_model_idx = GetDropdownChoiceIndex(kWhisperModelChoices,

 		kNumWhisperModelChoices, app_c_.whisper_model, kWhisperModelDefault);

-	whisper_model->SetSelection(model_idx);

+	whisper_model->SetSelection(whisper_model_idx);

 

     auto* whisper_button = static_cast<wxChoice*>(FindWindowById(ID_WHISPER_BUTTON));

 	whisper_button->SetSelection(button_idx);

@@ -1301,6 +1299,20 @@ void Frame::ApplyConfigToInputFields()
     unity_cols->AppendText(std::to_string(app_c_.cols));

 }

 

+void Frame::PopulateDynamicInputFields()

+{

+    whisper_->Init();

+    std::vector<std::string> mics;

+    if (whisper_->GetMics(mics)) {

+		std::vector<wxString> contents(mics.size());

+        auto* whisper_mic = static_cast<wxChoice*>(FindWindowById(ID_WHISPER_MIC));

+        for (int i = 0; i < std::min(mics.size(), kNumWhisperMicChoices); i++) {

+            contents[i] = mics[i];

+        }

+        whisper_mic->Set(contents);

+    }

+}

+

 void Frame::OnExit(wxCommandEvent& event)

 {

     OnAppStop(event);

@@ -1422,17 +1434,6 @@ void Frame::OnDumpMics(wxCommandEvent& event)
     Log(transcribe_out_, "{}\n", PythonWrapper::DumpMics());

 }

 

-void Frame::OnWhisperDumpMics(wxCommandEvent& event)

-{

-    whisper_->Init();

-    std::vector<std::string> mics;

-    whisper_->GetMics(mics);

-    Log(whisper_out_, "Microphones:\n");

-    for (int i = 0; i < mics.size(); i++) {

-        Log(whisper_out_, "  {}: {}\n", i, mics[i]);

-    }

-}

-

 bool GetUserPath(const std::string& raw, std::filesystem::path& clean,

     const std::string& err_prefix = "", bool must_exist = true) {

     clean = raw;

@@ -1902,7 +1903,7 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
     }

     int which_model = whisper_model_->GetSelection();

     if (which_model == wxNOT_FOUND) {

-        which_model = kModelDefault;

+        which_model = kWhisperModelDefault;

     }

     int chars_per_sync_idx = whisper_chars_per_sync_->GetSelection();

     if (chars_per_sync_idx == wxNOT_FOUND) {

@@ -1965,7 +1966,7 @@ void Frame::OnWhisperStart(wxCommandEvent& event) {
         return;

     }

 

-    app_c_.microphone = kMicChoices[which_mic].ToStdString();

+    app_c_.whisper_mic = which_mic;

     app_c_.language = kLangChoices[which_lang].ToStdString();

     app_c_.whisper_model = kWhisperModelChoices[which_model].ToStdString();

     app_c_.chars_per_sync = chars_per_sync;

diff --git a/GUI/GUI/GUI/Frame.h b/GUI/GUI/GUI/Frame.h
index ef65d7f..5d69679 100644
--- a/GUI/GUI/GUI/Frame.h
+++ b/GUI/GUI/GUI/Frame.h
@@ -87,6 +87,8 @@ private:
 

     // Initialize GUI input fields using `app_c_`.

     void ApplyConfigToInputFields();

+    // Populate dynamically-generated input fields, such as microphone lists.

+    void PopulateDynamicInputFields();

 

     void OnExit(wxCommandEvent& event);

     void OnNavbarTranscribe(wxCommandEvent& event);

@@ -95,7 +97,6 @@ private:
     void OnNavbarWhisper(wxCommandEvent& event);

     void OnSetupPython(wxCommandEvent& event);

     void OnDumpMics(wxCommandEvent& event);

-    void OnWhisperDumpMics(wxCommandEvent& event);

     void OnAppStart(wxCommandEvent& event);

     void OnAppStop(wxCommandEvent& event);

     void OnWhisperStart(wxCommandEvent& event);

diff --git a/GUI/GUI/GUI/WhisperCPP.cpp b/GUI/GUI/GUI/WhisperCPP.cpp
index 742f0a4..809415e 100644
--- a/GUI/GUI/GUI/WhisperCPP.cpp
+++ b/GUI/GUI/GUI/WhisperCPP.cpp
@@ -251,13 +251,12 @@ void WhisperCPP::Start(const AppConfig& c) {
 		return;

 	}

 

-	// TODO(yum) use app config to select mic

 	proc_ = new AppThread([&](AppThread* thd) {

 		Log(out_, "Transcription thread top\n");

 		run_ = true;

 

 		Whisper::iAudioCapture* mic_stream;

-		if (!OpenMic(1, mic_stream)) {

+		if (!OpenMic(c.whisper_mic, mic_stream)) {

 			return;

 		}

 		ScopeGuard mic_stream_cleanup([mic_stream]() { mic_stream->Release(); });

@@ -322,7 +321,8 @@ void WhisperCPP::Start(const AppConfig& c) {
 		Whisper::sFullParams wparams{};

 		context->fullDefaultParams(eSamplingStrategy::BeamSearch, &wparams);

 		wparams.language = Whisper::makeLanguageKey("en");  // TODO(yum) use config

-		wparams.n_max_text_ctx = 20;

+		// This must be set to keep memory usage from growing without bound.

+		wparams.n_max_text_ctx = 100;

 

 		wparams.new_segment_callback = [](iContext* context, uint32_t n_new, void* user_data) noexcept -> HRESULT {

 			wxTextCtrl* out = static_cast<wxTextCtrl*>(user_data);

@@ -346,12 +346,28 @@ void WhisperCPP::Start(const AppConfig& c) {
 			const int s0 = length.countSegments - n_new;

 			for (int i = s0; i < length.countSegments; i++) {

 				const sSegment& seg = segments[i];

-				Log(out, "{} ", seg.text);

+				bool is_metadata = false;

 				for (int j = 0; j < seg.countTokens; j++) {

 					const sToken& tok = tokens[seg.firstToken + j];

-					if (*tok.text == 0 || tok.text[0] == '[') {

+					if (tok.text[0] == '[') {

 						continue;

 					}

+					if (tok.text[0] == ' ' && (

+						tok.text[1] == '[' ||

+						tok.text[1] == '(')) {

+						if (tok.text[strlen(tok.text) - 1] == ']') {

+							continue;

+						}

+						is_metadata = true;

+						continue;

+					}

+					if (is_metadata && (

+						tok.text[strlen(tok.text) - 1] == ']' ||

+						tok.text[strlen(tok.text) - 1] == ')')) {

+						is_metadata = false;

+						continue;

+					}

+					Log(out, "{}", tok.text);

 				}

 			}

 			if (n_new) {

@@ -369,11 +385,12 @@ void WhisperCPP::Start(const AppConfig& c) {
 				Log(app->out_, "Exit transcription loop\n");

 				return S_FALSE;

 			}

+			// Sleeping here prevents the GUI from hanging.

+			// For some reason, printing is also required to prevent hanging.

 			static int i = 0;

-			if (++i % 10 == 0) {

+			if (++i % 20 == 0) {

 				Log(app->out_, "Spin {}\n", i);

 			}

-			// Sleeping here prevents the GUI from hanging.

 			wxThread::Sleep(10);

 			return S_OK;

 		};