feat: sound detection

README.md

@@ -35,7 +35,8 @@ transui             # same as record
 ```
 
 Bind `transui record` to a shortcut in your desktop environment. Transui does
-not register a global shortcut itself.
+not register a global shortcut itself. Recording stops automatically after two
+continuous seconds without detected speech.
 
 The **Hosted first, local fallback** setting tries the Hosted endpoint, then
 the Local endpoint when the first request fails. With it disabled, Transui uses

src/main.cpp

@@ -19,6 +19,8 @@ namespace {
 
 constexpr const char* kAppId = "io.github.transui";
 constexpr const char* kConfigGroup = "transui";
+constexpr gint64 kSilenceTimeoutUs = 2 * G_USEC_PER_SEC;
+constexpr double kSoundThresholdDb = -42.0;
 
 const SecretSchema kSecretSchema = {
     "io.github.transui.credentials", SECRET_SCHEMA_NONE,
@@ -43,7 +45,10 @@ struct App {
   GtkWidget* processing = nullptr;
   GtkWidget* elapsed_label = nullptr;
   guint elapsed_source = 0;
+  guint silence_source = 0;
+  guint level_watch = 0;
   guint elapsed_seconds = 0;
+  gint64 last_sound_us = 0;
   GstElement* pipeline = nullptr;
   std::string recording_path;
   std::atomic<bool> cancelled{false};
@@ -285,6 +290,55 @@ gboolean tick_elapsed(gpointer data) {
   return G_SOURCE_CONTINUE;
 }
 
+void stop_recording(GtkButton*, gpointer data);
+
+gboolean monitor_audio_level(GstBus*, GstMessage* message, gpointer data) {
+  if (GST_MESSAGE_TYPE(message) != GST_MESSAGE_ELEMENT) return G_SOURCE_CONTINUE;
+  const GstStructure* structure = gst_message_get_structure(message);
+  if (!structure || !gst_structure_has_name(structure, "level")) return G_SOURCE_CONTINUE;
+  const GValue* rms = gst_structure_get_value(structure, "rms");
+  if (!rms || !GST_VALUE_HOLDS_LIST(rms)) return G_SOURCE_CONTINUE;
+
+  auto* app = static_cast<App*>(data);
+  for (guint index = 0; index < gst_value_list_get_size(rms); ++index) {
+    const GValue* channel = gst_value_list_get_value(rms, index);
+    if (G_VALUE_HOLDS_DOUBLE(channel) && g_value_get_double(channel) > kSoundThresholdDb) {
+      app->last_sound_us = g_get_monotonic_time();
+      break;
+    }
+  }
+  return G_SOURCE_CONTINUE;
+}
+
+gboolean detect_silence(gpointer data) {
+  auto* app = static_cast<App*>(data);
+  if (!app->pipeline) {
+    app->silence_source = 0;
+    return G_SOURCE_REMOVE;
+  }
+  if (g_get_monotonic_time() - app->last_sound_us < kSilenceTimeoutUs)
+    return G_SOURCE_CONTINUE;
+
+  app->silence_source = 0;
+  stop_recording(nullptr, app);
+  return G_SOURCE_REMOVE;
+}
+
+void remove_recording_sources(App* app) {
+  if (app->elapsed_source) {
+    g_source_remove(app->elapsed_source);
+    app->elapsed_source = 0;
+  }
+  if (app->silence_source) {
+    g_source_remove(app->silence_source);
+    app->silence_source = 0;
+  }
+  if (app->level_watch) {
+    g_source_remove(app->level_watch);
+    app->level_watch = 0;
+  }
+}
+
 void remove_recording(App* app) {
   if (!app->recording_path.empty()) {
     g_remove(app->recording_path.c_str());
@@ -293,10 +347,7 @@ void remove_recording(App* app) {
 }
 
 void close_recorder(App* app) {
-  if (app->elapsed_source) {
+  remove_recording_sources(app);
-    g_source_remove(app->elapsed_source);
-    app->elapsed_source = 0;
-  }
   if (app->pipeline) {
     gst_element_set_state(app->pipeline, GST_STATE_NULL);
     gst_object_unref(app->pipeline);
@@ -506,10 +557,7 @@ void show_processing(App* app) {
 void stop_recording(GtkButton*, gpointer data) {
   auto* app = static_cast<App*>(data);
   if (!app->pipeline) return;
-  if (app->elapsed_source) {
+  remove_recording_sources(app);
-    g_source_remove(app->elapsed_source);
-    app->elapsed_source = 0;
-  }
   gst_element_send_event(app->pipeline, gst_event_new_eos());
   GstBus* bus = gst_element_get_bus(app->pipeline);
   GstMessage* message = gst_bus_timed_pop_filtered(bus, 3 * GST_SECOND,
@@ -537,10 +585,7 @@ void cancel_recording(GtkButton*, gpointer data) {
 void recorder_closed(GtkWidget*, gpointer data) {
   auto* app = static_cast<App*>(data);
   app->recorder = nullptr;
-  if (app->elapsed_source) {
+  remove_recording_sources(app);
-    g_source_remove(app->elapsed_source);
-    app->elapsed_source = 0;
-  }
   if (app->pipeline) {
     gst_element_set_state(app->pipeline, GST_STATE_NULL);
     gst_object_unref(app->pipeline);
@@ -562,7 +607,10 @@ void start_recording(App* app) {
   close(fd);
   app->recording_path = template_path;
   g_free(template_path);
-  gchar* pipeline_text = g_strdup_printf("autoaudiosrc ! audioconvert ! audioresample ! wavenc ! filesink location=\"%s\"", app->recording_path.c_str());
+  gchar* pipeline_text = g_strdup_printf(
+      "autoaudiosrc ! audioconvert ! audioresample ! "
+      "level interval=100000000 post-messages=true ! wavenc ! filesink location=\"%s\"",
+      app->recording_path.c_str());
   GError* error = nullptr;
   app->pipeline = gst_parse_launch(pipeline_text, &error);
   g_free(pipeline_text);
@@ -572,6 +620,9 @@ void start_recording(App* app) {
     remove_recording(app);
     return;
   }
+  GstBus* level_bus = gst_element_get_bus(app->pipeline);
+  app->level_watch = gst_bus_add_watch(level_bus, monitor_audio_level, app);
+  gst_object_unref(level_bus);
   if (gst_element_set_state(app->pipeline, GST_STATE_PLAYING) == GST_STATE_CHANGE_FAILURE) {
     show_error(nullptr, "Could not access an audio input.");
     close_recorder(app);
@@ -608,7 +659,9 @@ void start_recording(App* app) {
   gtk_box_pack_start(GTK_BOX(actions), cancel, false, false, 0);
   gtk_box_pack_start(GTK_BOX(actions), stop, false, false, 0);
   app->elapsed_seconds = 0;
+  app->last_sound_us = g_get_monotonic_time();
   app->elapsed_source = g_timeout_add_seconds(1, tick_elapsed, app);
+  app->silence_source = g_timeout_add(100, detect_silence, app);
   g_signal_connect(cancel, "clicked", G_CALLBACK(cancel_recording), app);
   g_signal_connect(stop, "clicked", G_CALLBACK(stop_recording), app);
   g_signal_connect(window, "destroy", G_CALLBACK(recorder_closed), app);
@@ -618,83 +671,36 @@ void start_recording(App* app) {
 
 void install_css() {
   const char* css = R"CSS(
-    window {
-      background-color: #181b1a;
-      color: #eef0e8;
-      font-family: Cantarell, sans-serif;
-    }
-    headerbar.transui-header {
-      background: #202421;
-      color: #eef0e8;
-      border-bottom: 1px solid #3b413c;
-      box-shadow: none;
-    }
     .settings-panel {
-      background-color: #202421;
+      padding: 12px 4px;
-      border: 1px solid #3b413c;
-      border-radius: 8px;
-      padding: 20px;
     }
     .field-label {
-      color: #aeb7ae;
       font-size: 12px;
       font-weight: 700;
+      opacity: 0.72;
     }
-    entry, combobox button {
-      background: #2a2f2b;
-      color: #eef0e8;
-      border: 1px solid #4a514b;
-      border-radius: 4px;
-      padding: 7px 9px;
-    }
-    entry:focus, combobox button:focus, button:focus {
-      border-color: #d4a44c;
-      box-shadow: 0 0 0 1px #d4a44c;
-    }
-    label { color: #d8ddd5; }
-    checkbutton { color: #d8ddd5; }
     .record-panel, .process-panel {
-      background-color: #202421;
-      border: 1px solid #3b413c;
-      border-radius: 12px;
       padding: 28px 36px;
     }
     .record-light {
-      background: #e64a3c;
+      background: #df493e;
-      border: 6px solid #552a25;
+      border: 6px solid alpha(#df493e, 0.22);
       border-radius: 50%;
-      box-shadow: 0 0 0 1px #f17a6f;
     }
     .record-state {
-      color: #f17a6f;
+      color: #c93e35;
       font-weight: 800;
       letter-spacing: 2px;
       font-size: 11px;
     }
     .elapsed {
-      color: #eef0e8;
       font-family: "DejaVu Sans Mono", monospace;
       font-size: 42px;
       font-weight: 700;
       letter-spacing: 1px;
     }
-    .state-title { color: #eef0e8; font-size: 17px; font-weight: 700; }
+    .state-title { font-size: 17px; font-weight: 700; }
-    .signal-spinner { color: #d4a44c; }
+    button.suggested-action { font-weight: 700; }
-    button {
-      background: #2a2f2b;
-      color: #eef0e8;
-      border: 1px solid #4a514b;
-      border-radius: 4px;
-      padding: 7px 12px;
-    }
-    button:hover { background: #343a35; }
-    button.suggested-action {
-      background: #d4a44c;
-      color: #181b1a;
-      border-color: #d4a44c;
-      font-weight: 800;
-    }
-    button.suggested-action:hover { background: #e1b760; }
   )CSS";
   GtkCssProvider* provider = gtk_css_provider_new();
   gtk_css_provider_load_from_data(provider, css, -1, nullptr);