Make min silence duration configurable in UI

4 files changed, 26 insertions, 0 deletions

diff --git a/GUI/GUI/GUI/Config.cpp b/GUI/GUI/GUI/Config.cpp
index 8d5cdbf..1fc1aee 100644
--- a/GUI/GUI/GUI/Config.cpp
+++ b/GUI/GUI/GUI/Config.cpp
@@ -85,6 +85,7 @@ AppConfig::AppConfig(wxTextCtrl* out)
 	enable_previews(true),

 	enable_lock_at_spawn(true),

 	gpu_idx(0),

+	min_silence_duration_ms(250),

 	keybind("ctrl+x"),

 

 	chars_per_sync(8),

@@ -127,6 +128,7 @@ bool AppConfig::Serialize(const std::filesystem::path& path) {
 	cm.Set("enable_previews", enable_previews);

 	cm.Set("enable_lock_at_spawn", enable_lock_at_spawn);

 	cm.Set("gpu_idx", gpu_idx);

+	cm.Set("min_silence_duration_ms", min_silence_duration_ms);

 	cm.Set("keybind", keybind);

 

 	cm.Set("chars_per_sync", chars_per_sync);

@@ -182,6 +184,7 @@ bool AppConfig::Deserialize(const std::filesystem::path& path) {
 	cm.Get("enable_previews", c.enable_previews);

 	cm.Get("enable_lock_at_spawn", c.enable_lock_at_spawn);

 	cm.Get("gpu_idx", c.gpu_idx);

+	cm.Get("min_silence_duration_ms", c.min_silence_duration_ms);

 	cm.Get("keybind", c.keybind);

 

 	cm.Get("chars_per_sync", c.chars_per_sync);

diff --git a/GUI/GUI/GUI/Config.h b/GUI/GUI/GUI/Config.h
index 6e82f14..808cf9e 100644
--- a/GUI/GUI/GUI/Config.h
+++ b/GUI/GUI/GUI/Config.h
@@ -71,6 +71,7 @@ public:
 	bool enable_previews;

 	bool enable_lock_at_spawn;

 	int gpu_idx;

+	int min_silence_duration_ms;

 	std::string keybind;

 

 	// Unity and transcription shared settings.

diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index a124ff1..d110a0c 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -75,6 +75,7 @@ namespace {
         ID_PY_APP_ROWS,

         ID_PY_APP_COLS,

         ID_PY_APP_GPU_IDX,

+        ID_PY_APP_MIN_SILENCE_DURATION_MS,

         ID_PY_APP_KEYBIND,

         ID_PY_APP_BROWSER_SRC_PORT,

         ID_PY_APP_COMMIT_FUZZ_THRESHOLD,

@@ -753,6 +754,15 @@ Frame::Frame()
 						"discrete GPU.");

 					py_app_gpu_idx_ = py_app_gpu_idx;

 

+					auto* py_app_min_silence_duration_ms = new wxTextCtrl(

+                        py_app_config_panel_pairs, ID_PY_APP_MIN_SILENCE_DURATION_MS,

+                        std::to_string(app_c_->min_silence_duration_ms), wxDefaultPosition,

+						wxDefaultSize, /*style=*/0);

+                    py_app_min_silence_duration_ms->SetToolTip(

+                        "The minimum duration, in milliseconds, of a silence "

+                        "used to segment speech.");

+					py_app_min_silence_duration_ms_ = py_app_min_silence_duration_ms;

+

 					auto* py_app_keybind = new wxTextCtrl(

 						py_app_config_panel_pairs, ID_PY_APP_KEYBIND,

 						app_c_->keybind, wxDefaultPosition,

@@ -838,6 +848,11 @@ Frame::Frame()
                         /*flags=*/wxEXPAND);

 

                     sizer->Add(new wxStaticText(py_app_config_panel_pairs,

+                        wxID_ANY, /*label=*/"Minimum silence duration (ms):"));

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

+                        /*flags=*/wxEXPAND);

+

+                    sizer->Add(new wxStaticText(py_app_config_panel_pairs,

                         wxID_ANY, /*label=*/"Browser source port:"));

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

                         /*flags=*/wxEXPAND);

@@ -1531,6 +1546,10 @@ void Frame::ApplyConfigToInputFields()
     py_app_gpu_idx->Clear();

     py_app_gpu_idx->AppendText(std::to_string(app_c_->gpu_idx));

 

+	auto* py_app_min_silence_duration_ms = static_cast<wxTextCtrl*>(FindWindowById(ID_PY_APP_MIN_SILENCE_DURATION_MS));

+	py_app_min_silence_duration_ms->Clear();

+	py_app_min_silence_duration_ms->AppendText(std::to_string(app_c_->min_silence_duration_ms));

+

     auto* py_app_enable_local_beep = static_cast<wxCheckBox*>(FindWindowById(ID_PY_APP_ENABLE_LOCAL_BEEP));

     py_app_enable_local_beep->SetValue(app_c_->enable_local_beep);

 

@@ -2284,6 +2303,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
 	ASSIGN_OR_RETURN_VOID(int, chars_per_sync, stoiInRange(transcribe_out_, kCharsPerSync[chars_per_sync_idx].ToStdString(), "chars_per_sync", 5, 24));

 	ASSIGN_OR_RETURN_VOID(int, bytes_per_char, stoiInRange(transcribe_out_, kBytesPerChar[bytes_per_char_idx].ToStdString(), "bytes_per_char", 1, 2));

 	ASSIGN_OR_RETURN_VOID(int, gpu_idx, stoiInRange(transcribe_out_, py_app_gpu_idx_->GetValue().ToStdString(), "gpu_idx", 0, 10));

+	ASSIGN_OR_RETURN_VOID(int, min_silence_duration_ms, stoiInRange(transcribe_out_, py_app_min_silence_duration_ms_->GetValue().ToStdString(), "min_silence_duration_ms", 50, 5000));

 	ASSIGN_OR_RETURN_VOID(int, browser_src_port, stoiInRange(transcribe_out_, py_app_browser_src_port_->GetValue().ToStdString(), "browser_src_port", 1024, 65535));

 

     std::string keybind = py_app_keybind_->GetValue().ToStdString();

@@ -2313,6 +2333,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
     app_c_->enable_previews = enable_previews;

     app_c_->enable_lock_at_spawn = enable_lock_at_spawn;

     app_c_->gpu_idx = gpu_idx;

+    app_c_->min_silence_duration_ms = min_silence_duration_ms;

     app_c_->keybind = keybind;

     app_c_->Serialize(AppConfig::kConfigPath);

 

diff --git a/GUI/GUI/GUI/Frame.h b/GUI/GUI/GUI/Frame.h
index d33db06..72ba6c4 100644
--- a/GUI/GUI/GUI/Frame.h
+++ b/GUI/GUI/GUI/Frame.h
@@ -38,6 +38,7 @@ private:
     wxTextCtrl* py_app_rows_;

     wxTextCtrl* py_app_cols_;

     wxTextCtrl* py_app_gpu_idx_;

+    wxTextCtrl* py_app_min_silence_duration_ms_;

     wxTextCtrl* py_app_keybind_;

     wxTextCtrl* py_app_browser_src_port_;

     wxTextCtrl* py_app_commit_fuzz_threshold_;