refactor: Update whisper model path handling in transcription filter

src/model-utils/model-downloader-ui.cpp

@@ -231,7 +231,9 @@ ModelDownloader::~ModelDownloader()
 		}
 		delete this->download_thread;
 	}
-	delete this->download_worker;
+	if (this->download_worker != nullptr) {
+		delete this->download_worker;
+	}
 }
 
 ModelDownloadWorker::~ModelDownloadWorker()

src/transcription-filter-callbacks.cpp

@@ -300,10 +300,7 @@ void enable_callback(void *data_, calldata_t *cd)
 		obs_log(gf_->log_level, "enable_callback: enable");
 		gf_->active = true;
 		reset_caption_state(gf_);
-		// get filter settings from gf_->context
-		obs_data_t *settings = obs_source_get_settings(gf_->context);
-		update_whisper_model(gf_, settings);
-		obs_data_release(settings);
+		update_whisper_model(gf_);
 	} else {
 		obs_log(gf_->log_level, "enable_callback: disable");
 		gf_->active = false;

src/transcription-filter.cpp

@@ -171,7 +171,7 @@ void transcription_filter_update(void *data, obs_data_t *s)
 	struct transcription_filter_data *gf =
 		static_cast<struct transcription_filter_data *>(data);
 
-	gf->log_level = LOG_INFO; //(int)obs_data_get_int(s, "log_level");
+	gf->log_level = (int)obs_data_get_int(s, "log_level");
 	gf->vad_enabled = obs_data_get_bool(s, "vad_enabled");
 	gf->log_words = obs_data_get_bool(s, "log_words");
 	gf->caption_to_stream = obs_data_get_bool(s, "caption_to_stream");
@@ -216,9 +216,9 @@ void transcription_filter_update(void *data, obs_data_t *s)
 			}
 		}
 	} else {
-		obs_log(LOG_INFO, "buffered_output disable");
+		obs_log(gf->log_level, "buffered_output disable");
 		if (gf->buffered_output) {
-			obs_log(LOG_INFO, "buffered_output currently enabled, disabling");
+			obs_log(gf->log_level, "buffered_output currently enabled, disabling");
 			if (gf->captions_monitor.isEnabled()) {
 				gf->captions_monitor.clear();
 				gf->captions_monitor.stopThread();
@@ -345,11 +345,11 @@ void transcription_filter_update(void *data, obs_data_t *s)
 		}
 	}
 
-	if (gf->initial_creation && obs_source_enabled(gf->context)) {
+	if (gf->initial_creation && gf->context != nullptr && obs_source_enabled(gf->context)) {
+		obs_log(LOG_INFO, "Initial filter creation and source enabled");
+
 		// source was enabled on creation
-		obs_data_t *settings = obs_source_get_settings(gf->context);
-		update_whisper_model(gf, settings);
-		obs_data_release(settings);
+		update_whisper_model(gf);
 		gf->active = true;
 		gf->initial_creation = false;
 	}
@@ -497,7 +497,7 @@ void transcription_filter_hide(void *data)
 
 void transcription_filter_defaults(obs_data_t *s)
 {
-	obs_log(LOG_INFO, "filter defaults");
+	obs_log(LOG_DEBUG, "filter defaults");
 
 	obs_data_set_default_bool(s, "buffered_output", false);
 	obs_data_set_default_int(s, "buffer_num_lines", 2);

src/whisper-utils/whisper-model-utils.cpp

@@ -9,10 +9,28 @@
 #include "plugin-support.h"
 #include "model-utils/model-downloader.h"
 
-void update_whisper_model(struct transcription_filter_data *gf, obs_data_t *s)
+void update_whisper_model(struct transcription_filter_data *gf)
 {
-	// update the whisper model path
+	if (gf->context == nullptr) {
+		obs_log(LOG_ERROR, "obs_source_t context is null");
+		return;
+	}
+
+	obs_data_t *s = obs_source_get_settings(gf->context);
+	if (s == nullptr) {
+		obs_log(LOG_ERROR, "obs_data_t settings is null");
+		return;
+	}
+
+	// Get settings from context
 	std::string new_model_path = obs_data_get_string(s, "whisper_model_path");
+	std::string external_model_file_path =
+		obs_data_get_string(s, "whisper_model_path_external");
+	const bool new_dtw_timestamps = obs_data_get_bool(s, "dtw_token_timestamps");
+	obs_data_release(s);
+
+	// update the whisper model path
+
 	const bool is_external_model = new_model_path.find("!!!external!!!") != std::string::npos;
 
 	char *silero_vad_model_file = obs_module_file("models/silero-vad/silero_vad.onnx");
@@ -73,8 +91,6 @@ void update_whisper_model(struct transcription_filter_data *gf, obs_data_t *s)
 			}
 		} else {
 			// new model is external file, get file location from file property
-			std::string external_model_file_path =
-				obs_data_get_string(s, "whisper_model_path_external");
 			if (external_model_file_path.empty()) {
 				obs_log(LOG_WARNING, "External model file path is empty");
 			} else {
@@ -98,13 +114,11 @@ void update_whisper_model(struct transcription_filter_data *gf, obs_data_t *s)
 			gf->whisper_model_path.c_str(), new_model_path.c_str());
 	}
 
-	const bool new_dtw_timestamps = obs_data_get_bool(s, "dtw_token_timestamps");
-
 	if (new_dtw_timestamps != gf->enable_token_ts_dtw) {
 		// dtw_token_timestamps changed
 		obs_log(gf->log_level, "dtw_token_timestamps changed from %d to %d",
 			gf->enable_token_ts_dtw, new_dtw_timestamps);
-		gf->enable_token_ts_dtw = obs_data_get_bool(s, "dtw_token_timestamps");
+		gf->enable_token_ts_dtw = new_dtw_timestamps;
 		shutdown_whisper_thread(gf);
 		start_whisper_thread_with_path(gf, gf->whisper_model_path,
 					       silero_vad_model_file_str.c_str());

src/whisper-utils/whisper-model-utils.h

@@ -5,6 +5,6 @@
 
 #include "transcription-filter-data.h"
 
-void update_whisper_model(struct transcription_filter_data *gf, obs_data_t *s);
+void update_whisper_model(struct transcription_filter_data *gf);
 
 #endif // WHISPER_MODEL_UTILS_H

src/whisper-utils/whisper-processing.cpp

@@ -476,7 +476,7 @@ void whisper_loop(void *data)
 	struct transcription_filter_data *gf =
 		static_cast<struct transcription_filter_data *>(data);
 
-	obs_log(LOG_INFO, "starting whisper thread");
+	obs_log(gf->log_level, "Starting whisper thread");
 
 	vad_state current_vad_state = {false, 0, 0};
 	// 500 ms worth of audio is needed for VAD segmentation
@@ -511,5 +511,5 @@ void whisper_loop(void *data)
 		gf->wshiper_thread_cv.wait_for(lock, std::chrono::milliseconds(50));
 	}
 
-	obs_log(LOG_INFO, "exiting whisper thread");
+	obs_log(gf->log_level, "Exiting whisper thread");
 }