Playht delete circular buffer with mutex check (#123)

Signed-off-by: root <root@af6633a5cfe8>
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_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/playht_glue.cpp

@@ -959,8 +959,11 @@ 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;