support playht3.0 (#119)

* support playht3.0

* update language

* wip

* update top_p and repetition_penalty

---------

Co-authored-by: root <root@af6633a5cfe8>

mod_audio_fork/lws_glue.cpp

@@ -557,7 +557,7 @@ namespace {
       tech_pvt->streamingPreBuffer = nullptr;
     }
 
-    if (nullptr == tech_pvt->pVecMarksInInventory) {
+    if (tech_pvt->pVecMarksInInventory) {
       delete static_cast<std::deque<std::string>*>(tech_pvt->pVecMarksInInventory);
       tech_pvt->pVecMarksInInventory = nullptr;
       delete static_cast<std::deque<std::string>*>(tech_pvt->pVecMarksInUse);
@@ -979,7 +979,7 @@ extern "C" {
         std::deque<std::string>* pVecInventory = nullptr;
         std::deque<std::string>* pVecInUse = nullptr;
         std::deque<std::string>* pVecCleared = nullptr;
-        if (nullptr != tech_pvt->pVecMarksInUse) {
+        if (tech_pvt->pVecMarksInInventory && tech_pvt->pVecMarksInUse && tech_pvt->pVecMarksCleared) {
           pVecInventory = static_cast<std::deque<std::string>*>(tech_pvt->pVecMarksInInventory);
           pVecInUse = static_cast<std::deque<std::string>*>(tech_pvt->pVecMarksInUse);
           pVecCleared = static_cast<std::deque<std::string>*>(tech_pvt->pVecMarksCleared);
@@ -1032,7 +1032,7 @@ extern "C" {
 
           if (samplesToCopy > 0) {
             vector_add(fp, data, rframe->samples);
-          } else if (pVecInventory != nullptr && pVecInventory->size()) {
+          } else if (pVecInventory && pVecInventory->size()) {
             // no bidirectional audio to dub but still have some mark in inventory, send them now
             auto name = pVecInventory->front();
             pVecInventory->pop_front();

mod_azure_transcribe/azure_transcribe_glue.cpp

@@ -72,10 +72,7 @@ public:
 		if (switch_true(switch_channel_get_variable(channel, "AZURE_USE_OUTPUT_FORMAT_DETAILED"))) {
 			speechConfig->SetOutputFormat(OutputFormat::Detailed);
 		}
-		if (nullptr != endpointId) {
-      switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(psession), SWITCH_LOG_DEBUG, "setting endpoint id: %s\n", endpointId);
-			speechConfig->SetEndpointId(endpointId);
-		}
+
 		if (!sdkInitialized && sdkLog) {
 			sdkInitialized = true;
 			speechConfig->SetProperty(PropertyId::Speech_LogFilename, sdkLog);
@@ -104,11 +101,31 @@ public:
 				languages.push_back( alt_langs[i]);
         switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(psession), SWITCH_LOG_DEBUG, "added alternative lang %s\n", alt_langs[i]);
       }
-			auto autoDetectSourceLanguageConfig = AutoDetectSourceLanguageConfig::FromLanguages(languages);
+
+			std::vector<std::shared_ptr<SourceLanguageConfig>> sourceLanguageConfigs;
+			for (const auto& language : languages) {
+					std::shared_ptr<SourceLanguageConfig> sourceLanguageConfig;
+					
+					if (endpointId != nullptr) {
+							sourceLanguageConfig = SourceLanguageConfig::FromLanguage(language, endpointId);
+					} else {
+							sourceLanguageConfig = SourceLanguageConfig::FromLanguage(language);
+					}
+
+					sourceLanguageConfigs.push_back(sourceLanguageConfig);
+			}
+
+			// Create AutoDetectSourceLanguageConfig from SourceLanguageConfigs
+			auto autoDetectSourceLanguageConfig = AutoDetectSourceLanguageConfig::FromSourceLanguageConfigs(sourceLanguageConfigs);
 			m_recognizer = SpeechRecognizer::FromConfig(speechConfig, autoDetectSourceLanguageConfig, audioConfig);
     }
 		else {
-			auto sourceLanguageConfig = SourceLanguageConfig::FromLanguage(lang);
+			std::shared_ptr<SourceLanguageConfig> sourceLanguageConfig;
+			if (endpointId != nullptr) {
+					sourceLanguageConfig = SourceLanguageConfig::FromLanguage(lang, endpointId);
+			} else {
+					sourceLanguageConfig = SourceLanguageConfig::FromLanguage(lang);
+			}
 			m_recognizer = SpeechRecognizer::FromConfig(speechConfig, sourceLanguageConfig, audioConfig);
 		}
 
@@ -487,20 +504,27 @@ extern "C" {
 		auto read_codec = switch_core_session_get_read_codec(session);
 		uint32_t sampleRate = read_codec->implementation->actual_samples_per_second;
 		if (bug) {
-			struct cap_cb* existing_cb = (struct cap_cb*) switch_core_media_bug_get_user_data(bug);
-			GStreamer* existing_streamer = (GStreamer*) existing_cb->streamer;
-			existing_cb->is_keep_alive = 0;
-			if (!existing_streamer->hasConfigurationChanged(channels, lang, interim, sampleRate, region, subscriptionKey)) {
-				switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Reuse active azure connection.\n");
+			try {
+				struct cap_cb* existing_cb = (struct cap_cb*) switch_core_media_bug_get_user_data(bug);
+				GStreamer* existing_streamer = (GStreamer*) existing_cb->streamer;
+				existing_cb->is_keep_alive = 0;
+				if (!existing_streamer->hasConfigurationChanged(channels, lang, interim, sampleRate, region, subscriptionKey)) {
+					switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Reuse active azure connection.\n");
+					return SWITCH_STATUS_SUCCESS;
+				}
+				switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Azure configuration is changed, destroy old and create new azure connection\n");
+				reaper(existing_cb);
+				streamer =  new GStreamer(sessionId, bugname, channels, lang, interim, sampleRate, region, subscriptionKey, responseHandler);
+				if (!existing_cb->vad) streamer->connect();
+				existing_cb->streamer = streamer;
+				*ppUserData = existing_cb;
 				return SWITCH_STATUS_SUCCESS;
+			} catch (std::exception& e) {
+				switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_ERROR, "%s: Error initializing gstreamer: %s.\n", 
+					switch_channel_get_name(channel), e.what());
+				return SWITCH_STATUS_FALSE;
 			}
-			switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Azure configuration is changed, destroy old and create new azure connection\n");
-			reaper(existing_cb);
-			streamer =  new GStreamer(sessionId, bugname, channels, lang, interim, sampleRate, region, subscriptionKey, responseHandler);
-			if (!existing_cb->vad) streamer->connect();
-			existing_cb->streamer = streamer;
-			*ppUserData = existing_cb;
-			return SWITCH_STATUS_SUCCESS;
+			
 		}
 		int err;
 		switch_threadattr_t *thd_attr = NULL;

mod_elevenlabs_tts/README.md

@@ -1,11 +1,11 @@
 # mod_google_tts
 
-A Freeswitch module that allows Google Text-to-Speech API to be used as a tts provider.
+A Freeswitch module that allows Eleven Labs' Text-to-Speech API to be used as a tts provider.
 
 ## API
 
 ### Commands
-This freeswitch module does not add any new commands, per se.  Rather, it integrates into the Freeswitch TTS interface such that it is invoked when an application uses the mod_dptools `speak` command with a tts engine of `google_tts` and a voice equal to the language code associated to one of the [supported Wavenet voices](https://cloud.google.com/text-to-speech/docs/voices)
+This freeswitch module does not add any new commands, per se.  Rather, it integrates into the Freeswitch TTS interface such that it is invoked when an application uses the mod_dptools `speak` command with a tts engine of `elevenlabs` and a voice equal to the language code associated to one of the [supported Eleven Labs voices](https://elevenlabs.io/docs/api-reference/query-library)
 
 ### Events
 None.
@@ -13,11 +13,19 @@ None.
 ## Usage
 When using [drachtio-fsrmf](https://www.npmjs.com/package/drachtio-fsmrf), you can access this functionality via the speak method on the 'endpoint' object.
 ```js
-ep.speak({
-    ttsEngine: 'google_tts',
-    voice: 'en-GB-Wavenet-A',
-    text: 'This aggression will not stand'
-  });
+var text = "Hello World";
+await endpoint.speak({
+    "ttsEngine": 'elevenlabs',
+    "voice": "W9OIfHh5DtdYiZUcFiql",
+    "text": `{use_speaker_boost=1,optimize_streaming_latency=4,style=0.5,stability=0.5,similarity_boost=0.75,api_key=XXYYZZ,model_id=eleven_turbo_v2}${text}`,
+});
 ```
-## Examples
-[google_tts.js](../../examples/google_tts.js)
+## Options
+
+Documentation on these options can be found in Eleven Labs API docs: [Voice Settings](https://elevenlabs.io/docs/speech-synthesis/voice-settings)
+
+- use_speaker_boost
+- optimize_streaming_latency
+- style
+- stability
+- similarity_boost

mod_google_transcribe/google_glue_v1.cpp

@@ -233,6 +233,10 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
     auto speech_event_type = response.speech_event_type();
     if (response.has_error()) {
       Status status = response.error();
+      //error 11 is handled in finished session, avoid sending jambonz_transcribe::error event for this here.
+      if (11 == status.code()) {
+        continue;
+      }
       switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "grpc_read_thread: error %s (%d)\n", status.message().c_str(), status.code()) ;
       cJSON* json = cJSON_CreateObject();
       cJSON_AddStringToObject(json, "type", "error");

mod_playht_tts/mod_playht_tts.c

@@ -14,6 +14,10 @@ static void clearPlayht(playht_t* p, int freeAll) {
   if (p->seed) free(p->seed);
   if (p->temperature) free(p->temperature);
   if (p->voice_engine) free(p->voice_engine);
+  if (p->synthesize_url) free(p->synthesize_url);
+  if (p->language) free(p->language);
+  if (p->top_p) free(p->top_p);
+  if (p->repetition_penalty) free(p->repetition_penalty);
   if (p->emotion) free(p->emotion);
   if (p->voice_guidance) free(p->voice_guidance);
   if (p->style_guidance) free(p->style_guidance);
@@ -36,6 +40,10 @@ static void clearPlayht(playht_t* p, int freeAll) {
   p->seed = NULL;
   p->temperature = NULL;
   p->voice_engine = NULL;
+  p->synthesize_url = NULL;
+  p->language = NULL;
+  p->top_p = NULL;
+  p->repetition_penalty = NULL;
   p->emotion = NULL;
   p->voice_guidance = NULL;
   p->style_guidance = NULL;
@@ -154,6 +162,18 @@ static void p_text_param_tts(switch_speech_handle_t *sh, char *param, const char
   } else if (0 == strcmp(param, "voice_engine")) {
     if (p->voice_engine) free(p->voice_engine);
     p->voice_engine = strdup(val);
+  } else if (0 == strcmp(param, "synthesize_url")) {
+    if (p->synthesize_url) free(p->synthesize_url);
+    p->synthesize_url = strdup(val);
+  } else if (0 == strcmp(param, "language")) {
+    if (p->language) free(p->language);
+    p->language = strdup(val);
+  } else if (0 == strcmp(param, "top_p")) {
+    if (p->top_p) free(p->top_p);
+    p->top_p = strdup(val);
+  } else if (0 == strcmp(param, "repetition_penalty")) {
+    if (p->repetition_penalty) free(p->repetition_penalty);
+    p->repetition_penalty = strdup(val);
   } else if (0 == strcmp(param, "emotion")) {
     if (p->emotion) free(p->emotion);
     p->emotion = strdup(val);

mod_playht_tts/mod_playht_tts.h

@@ -11,10 +11,14 @@ typedef struct playht_data {
   char *seed;
   char *temperature;
   char *voice_engine;
+  char *synthesize_url;
+  char *language;
   char *emotion;
   char *voice_guidance;
   char *style_guidance;
   char *text_guidance;
+  char *top_p;
+  char *repetition_penalty;
 
   /* result data */
   long response_code;

mod_playht_tts/playht_glue.cpp

@@ -780,7 +780,7 @@ extern "C" {
     }
 
     /* format url*/
-    std::string url = "https://api.play.ht/api/v2/tts/stream";
+    std::string url = p->synthesize_url;
 
     /* create the JSON body */
     cJSON * jResult = cJSON_CreateObject();
@@ -800,9 +800,6 @@ extern "C" {
         cJSON_AddNumberToObject(jResult, "speed", val);
       }
     }
-    if (p->seed) {
-      cJSON_AddNumberToObject(jResult, "seed", atoi(p->seed));
-    }
     if (p->temperature) {
       cJSON_AddNumberToObject(jResult, "temperature", std::strtof(p->temperature, nullptr));
     }
@@ -818,6 +815,15 @@ extern "C" {
     if (p->text_guidance) {
       cJSON_AddNumberToObject(jResult, "text_guidance", atoi(p->text_guidance));
     }
+    if (strcmp(p->voice_engine, "Play3.0") == 0 && p->language) {
+      cJSON_AddStringToObject(jResult, "language", p->language);
+    }
+    if (p->repetition_penalty) {
+      cJSON_AddNumberToObject(jResult, "repetition_penalty", std::strtof(p->repetition_penalty, nullptr));
+    }
+    if (p->top_p) {
+      cJSON_AddNumberToObject(jResult, "top_p", std::strtof(p->top_p, nullptr));
+    }
     char *json = cJSON_PrintUnformatted(jResult);
 
     cJSON_Delete(jResult);
@@ -889,9 +895,12 @@ extern "C" {
 
     /* call this function to close a socket */
     curl_easy_setopt(easy, CURLOPT_CLOSESOCKETFUNCTION, close_socket);
+    // Play3.0 voice engine doesn't need authorization
+    if (strcmp(p->voice_engine, "Play3.0") != 0) {
+      conn->hdr_list = curl_slist_append(conn->hdr_list, api_key_stream.str().c_str());
+      conn->hdr_list = curl_slist_append(conn->hdr_list, user_id_stream.str().c_str());
+    }
 
-    conn->hdr_list = curl_slist_append(conn->hdr_list, api_key_stream.str().c_str());
-    conn->hdr_list = curl_slist_append(conn->hdr_list, user_id_stream.str().c_str());
     conn->hdr_list = curl_slist_append(conn->hdr_list, "Accept: audio/mpeg");
     conn->hdr_list = curl_slist_append(conn->hdr_list, "Content-Type: application/json");
     curl_easy_setopt(easy, CURLOPT_HTTPHEADER, conn->hdr_list);
@@ -959,8 +968,11 @@ extern "C" {
     switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "playht_speech_flush_tts, download complete? %s\n", download_complete ? "yes" : "no") ;
     ConnInfo_t *conn = (ConnInfo_t *) p->conn;
     CircularBuffer_t *cBuffer = (CircularBuffer_t *) p->circularBuffer;
+    // In multi threads, only delete the circular buffer when write and read buffer action finished using it.
+    switch_mutex_lock(p->mutex);
     delete cBuffer;
     p->circularBuffer = nullptr ;
+    switch_mutex_unlock(p->mutex);
 
     if (conn) {
       conn->flushed = true;