feat: remove background cam every time

main/application.cc

@@ -81,6 +81,7 @@ void Application::Initialize() {
         xEventGroupSetBits(event_group_, MAIN_EVENT_WAKE_WORD_DETECTED);
     };
     callbacks.on_vad_change = [this](bool speaking) {
+        vad_speaking_.store(speaking);
         xEventGroupSetBits(event_group_, MAIN_EVENT_VAD_CHANGE);
     };
     audio_service_.SetCallbacks(callbacks);
@@ -233,6 +234,13 @@ void Application::Run() {
             if (GetDeviceState() == kDeviceStateListening) {
                 auto led = Board::GetInstance().GetLed();
                 led->OnStateChanged();
+
+                if (vad_speaking_.load() && vision_text_mode_enabled_.load() &&
+                    !vision_frame_sent_for_current_listen_.exchange(true)) {
+                    if (!SendCurrentVisionFrame()) {
+                        vision_frame_sent_for_current_listen_.store(false);
+                    }
+                }
             }
         }
 
@@ -674,11 +682,13 @@ void Application::DismissAlert() {
 
 void Application::ToggleChatState() {
     vision_text_mode_enabled_.store(false);
+    vision_frame_sent_for_current_listen_.store(false);
     xEventGroupSetBits(event_group_, MAIN_EVENT_TOGGLE_CHAT);
 }
 
 void Application::ToggleChatStateWithVision() {
     vision_text_mode_enabled_.store(true);
+    vision_frame_sent_for_current_listen_.store(false);
     xEventGroupSetBits(event_group_, MAIN_EVENT_TOGGLE_CHAT);
 }
 
@@ -688,6 +698,7 @@ bool Application::IsVisionTextModeEnabled() const {
 
 void Application::StartListening() {
     vision_text_mode_enabled_.store(false);
+    vision_frame_sent_for_current_listen_.store(false);
     xEventGroupSetBits(event_group_, MAIN_EVENT_START_LISTENING);
 }
 
@@ -697,9 +708,6 @@ void Application::StopListening() {
 
 void Application::HandleToggleChatEvent() {
     auto state = GetDeviceState();
-    if (state != kDeviceStateIdle) {
-        vision_text_mode_enabled_.store(false);
-    }
 
     if (state == kDeviceStateActivating) {
         SetDeviceState(kDeviceStateIdle);
@@ -896,6 +904,7 @@ void Application::HandleStateChangedEvent() {
     switch (new_state) {
         case kDeviceStateUnknown:
         case kDeviceStateIdle:
+            vision_frame_sent_for_current_listen_.store(false);
             display->SetStatus(Lang::Strings::STANDBY);
             display->ClearChatMessages();  // Clear messages first
             display->SetEmotion("neutral"); // Then set emotion (wechat mode checks child count)
@@ -908,6 +917,8 @@ void Application::HandleStateChangedEvent() {
             display->SetChatMessage("system", "");
             break;
         case kDeviceStateListening:
+            vad_speaking_.store(false);
+            vision_frame_sent_for_current_listen_.store(false);
             display->SetStatus(Lang::Strings::LISTENING);
             display->SetEmotion("neutral");
 
@@ -919,9 +930,6 @@ void Application::HandleStateChangedEvent() {
                     audio_service_.WaitForPlaybackQueueEmpty();
                 }
                 
-                if (vision_text_mode_enabled_.load()) {
-                    SendCurrentVisionFrame();
-                }
                 // Send the start listening command
                 protocol_->SendStartListening(listening_mode_);
                 audio_service_.EnableVoiceProcessing(true);
@@ -961,24 +969,25 @@ void Application::HandleStateChangedEvent() {
     }
 }
 
-void Application::SendCurrentVisionFrame() {
+bool Application::SendCurrentVisionFrame() {
     if (!protocol_ || !protocol_->IsAudioChannelOpened()) {
-        return;
+        return false;
     }
 
     auto camera = Board::GetInstance().GetCamera();
     if (camera == nullptr) {
-        return;
+        return false;
     }
 
     std::string jpeg_data;
     if (!camera->CaptureToJpeg(jpeg_data, true)) {
         ESP_LOGW(TAG, "Failed to capture vision frame");
-        return;
+        return false;
     }
 
     protocol_->SendVisionFrame(jpeg_data);
     ESP_LOGI(TAG, "Sent vision frame, size=%u bytes", static_cast<unsigned>(jpeg_data.size()));
+    return true;
 }
 
 void Application::Schedule(std::function<void()>&& callback) {
@@ -999,6 +1008,8 @@ void Application::AbortSpeaking(AbortReason reason) {
 
 void Application::SetListeningMode(ListeningMode mode) {
     listening_mode_ = mode;
+    vad_speaking_.store(false);
+    vision_frame_sent_for_current_listen_.store(false);
     SetDeviceState(kDeviceStateListening);
 }
 

main/application.h

@@ -148,6 +148,8 @@ private:
     bool assets_version_checked_ = false;
     bool play_popup_on_listening_ = false;  // Flag to play popup sound after state changes to listening
     std::atomic<bool> vision_text_mode_enabled_ = false;
+    std::atomic<bool> vad_speaking_ = false;
+    std::atomic<bool> vision_frame_sent_for_current_listen_ = false;
     int clock_ticks_ = 0;
     TaskHandle_t activation_task_handle_ = nullptr;
 
@@ -163,7 +165,7 @@ private:
     void HandleWakeWordDetectedEvent();
     void ContinueOpenAudioChannel(ListeningMode mode);
     void ContinueWakeWordInvoke(const std::string& wake_word);
-    void SendCurrentVisionFrame();
+    bool SendCurrentVisionFrame();
 
     // Activation task (runs in background)
     void ActivationTask();

main/boards/m5stack-core-s3/m5stack_core_s3.cc

@@ -344,21 +344,23 @@ private:
                 vTaskDelay(pdMS_TO_TICKS(BACKGROUND_VISION_INITIAL_DELAY_MS));
 
                 while (true) {
+                    if (!Application::GetInstance().IsVisionTextModeEnabled()) {
+                        vTaskDelay(pdMS_TO_TICKS(BACKGROUND_VISION_SAMPLE_INTERVAL_MS));
+                        continue;
+                    }
+
                     if (board->camera_ == nullptr) {
                         vTaskDelay(pdMS_TO_TICKS(BACKGROUND_VISION_SAMPLE_INTERVAL_MS));
                         continue;
                     }
 
-                    bool capture_ok = Application::GetInstance().IsVisionTextModeEnabled()
+                    if (board->camera_->Capture()) {
-                                          ? board->camera_->Capture()
-                                          : board->camera_->CaptureBackground();
-                    if (capture_ok) {
                         if (!has_logged_success) {
-                            ESP_LOGI(TAG, "Background vision sampler started");
+                            ESP_LOGI(TAG, "Vision preview sampler started");
                             has_logged_success = true;
                         }
                     } else if (!has_logged_failure) {
-                        ESP_LOGW(TAG, "Background vision sampler is waiting for camera");
+                        ESP_LOGW(TAG, "Vision preview sampler is waiting for camera");
                         has_logged_failure = true;
                     }