elevenlabs to use http 1.1 (#57)

* elevenlabs to use http 1.1

Signed-off-by: Hoan HL <quan.luuhoang8@gmail.com>

* update the rest of mod tts

Signed-off-by: Hoan HL <quan.luuhoang8@gmail.com>

* add env variable for disabling http2

Signed-off-by: Hoan HL <quan.luuhoang8@gmail.com>

---------

Signed-off-by: Hoan HL <quan.luuhoang8@gmail.com>

mod_audio_fork/audio_pipe.cpp

@@ -480,7 +480,7 @@ AudioPipe::AudioPipe(const char* uuid, const char* host, unsigned int port, cons
 }
 AudioPipe::~AudioPipe() {
   if (m_audio_buffer) delete [] m_audio_buffer;
-  if (m_recv_buf) delete [] m_recv_buf;
+  if (m_recv_buf) free(m_recv_buf);
 }
 
 void AudioPipe::connect(void) {

mod_azure_tts/azure_glue.cpp

@@ -135,23 +135,12 @@ extern "C" {
       return SWITCH_STATUS_FALSE;
     }
 
-    if (a->session_id) {
+    if (a->rate != 8000 /*Hz*/) {
       int err;
-      switch_codec_implementation_t read_impl;
-      
-      /* lock and unlock session */
-      switch_core_session_t *psession = switch_core_session_locate(a->session_id);
-      switch_core_session_get_read_impl(psession, &read_impl);
-      switch_core_session_rwunlock(psession);
-
-      uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
-      a->samples_rate = samples_per_second;
-      if (samples_per_second != 8000 /*Hz*/) {
-        a->resampler = speex_resampler_init(1, 8000, samples_per_second, SWITCH_RESAMPLE_QUALITY, &err);
-        if (0 != err) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
-          return SWITCH_STATUS_FALSE;
-        }
+      a->resampler = speex_resampler_init(1, 8000, a->rate, SWITCH_RESAMPLE_QUALITY, &err);
+      if (0 != err) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
+        return SWITCH_STATUS_FALSE;
       }
     }
 
@@ -276,9 +265,7 @@ extern "C" {
       return SWITCH_STATUS_SUCCESS;
     }
     // azure returned 8000hz 16 bit data, we have to take enough data based on call sample rate.
-    size_t size = a->samples_rate ?
-      std::min((*datalen/(2 * a->samples_rate / 8000)), bufSize) :
-      std::min((*datalen/2), bufSize);
+    size_t size = std::min((*datalen/(2 * a->rate / 8000)), bufSize);
     pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
     cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
     switch_mutex_unlock(a->mutex);

mod_azure_tts/mod_azure_tts.c

@@ -82,7 +82,6 @@ static switch_status_t a_speech_feed_tts(switch_speech_handle_t *sh, char *text,
   a->draining = 0;
   a->reads = 0;
   a->flushed = 0;
-  a->samples_rate = 0;
 
   return azure_speech_feed_tts(a, text, flags);
 }

mod_azure_tts/mod_azure_tts.h

@@ -25,7 +25,6 @@ typedef struct azure_data {
   int reads;
   int cache_audio;
   int flushed;
-  uint32_t samples_rate;
 
   void *startTime;
 

mod_deepgram_tts/deepgram_glue.cpp

@@ -764,7 +764,9 @@ extern "C" {
     conn->file = d->file;
     conn->body = json;
     conn->flushed = false;
-    
+    conn->has_last_byte = false;
+    conn->last_byte = 0;
+
 
     d->circularBuffer = (void *) new CircularBuffer_t(BUFFER_GROW_SIZE);
     // Always use deepgram at rate 8000 for helping cache audio from jambonz.
@@ -803,7 +805,10 @@ extern "C" {
     curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
     //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
 
-    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+    // libcurl adding random byte to the response body that creates white noise to audio file
+    // https://github.com/curl/curl/issues/10525
+    const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
     rc = curl_multi_add_handle(global.multi, conn->easy);
     mcode_test("new_conn: curl_multi_add_handle", rc);

mod_dub/ap_http.cpp

@@ -136,7 +136,10 @@ void AudioProducerHttp::start(std::function<void(bool, const std::string&)> call
   /* call this function to close a socket */
   curl_easy_setopt(_easy, CURLOPT_CLOSESOCKETFUNCTION, &AudioProducerHttp::close_socket);
 
-  curl_easy_setopt(_easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+  // libcurl adding random byte to the response body that creates white noise to audio file
+  // https://github.com/curl/curl/issues/10525
+  const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+  curl_easy_setopt(_easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
   /* keep the speed down so we don't have to buffer large amounts*/
   curl_easy_setopt(_easy, CURLOPT_MAX_RECV_SPEED_LARGE, (curl_off_t)31415);

mod_elevenlabs_tts/elevenlabs_glue.cpp

@@ -841,7 +841,6 @@ extern "C" {
     CURL* easy = createEasyHandle();
 
     el->conn = (void *) conn ;
-    el->sample_rate = 0;
     conn->elevenlabs = el;
     conn->easy = easy;
     conn->global = &global;
@@ -852,20 +851,12 @@ extern "C" {
 
     el->circularBuffer = (void *) new CircularBuffer_t(8192);
 
-    if (el->session_id) {
+    if (el->rate != 8000 /*Hz*/) {
       int err;
-      switch_codec_implementation_t read_impl;
-      switch_core_session_t *psession = switch_core_session_locate(el->session_id);
-      switch_core_session_get_read_impl(psession, &read_impl);
-      uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
-      el->sample_rate = samples_per_second;
-      // elevenlabs output is PCMU 8000
-      if (samples_per_second != 8000 /*Hz*/) {
-        el->resampler = speex_resampler_init(1, 8000, samples_per_second, SWITCH_RESAMPLE_QUALITY, &err);
-        if (0 != err) {
-          switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
-          return SWITCH_STATUS_FALSE;
-        }
+      el->resampler = speex_resampler_init(1, 8000, el->rate, SWITCH_RESAMPLE_QUALITY, &err);
+      if (0 != err) {
+        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error initializing resampler: %s.\n", speex_resampler_strerror(err));
+        return SWITCH_STATUS_FALSE;
       }
     }
 
@@ -895,7 +886,10 @@ extern "C" {
     curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
     //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
 
-    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+    // libcurl adding random byte to the response body that creates white noise to audio file
+    // https://github.com/curl/curl/issues/10525
+    const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
     rc = curl_multi_add_handle(global.multi, conn->easy);
     mcode_test("new_conn: curl_multi_add_handle", rc);
@@ -939,9 +933,7 @@ extern "C" {
         switch_mutex_unlock(el->mutex);
         return SWITCH_STATUS_SUCCESS;
       }
-      size_t size = el->sample_rate ?
-        std::min((*datalen/(2 * el->sample_rate / 8000)), cBuffer->size()) :
-        std::min((*datalen/2), cBuffer->size());
+      size_t size = std::min((*datalen/(2 * el->rate / 8000)), cBuffer->size());
       pcm_data.insert(pcm_data.end(), cBuffer->begin(), cBuffer->begin() + size);
       cBuffer->erase(cBuffer->begin(), cBuffer->begin() + size);
       switch_mutex_unlock(el->mutex);

mod_elevenlabs_tts/mod_elevenlabs_tts.h

@@ -31,7 +31,6 @@ struct elevenlabs_data {
   char *cache_filename;
 
 	int rate;
-  uint32_t sample_rate;
 
   void *conn;
 	FILE *file;

mod_google_transcribe/google_glue_v1.cpp

@@ -236,6 +236,7 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
       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");
+      cJSON_AddStringToObject(json, "error_cause", "stream_response");
       cJSON_AddStringToObject(json, "error", status.message().c_str());
       char* jsonString = cJSON_PrintUnformatted(json);
       cb->responseHandler(session, jsonString, cb->bugname);
@@ -339,9 +340,10 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
           cb->responseHandler(session, "no_audio", cb->bugname);
         }
       }
-      else {
+      else if (status.error_code() != 0) {
         cJSON* json = cJSON_CreateObject();
         cJSON_AddStringToObject(json, "type", "error");
+        cJSON_AddStringToObject(json, "error_cause", "stream_close");
         cJSON_AddItemToObject(json, "error_code", cJSON_CreateNumber(status.error_code()));
         cJSON_AddStringToObject(json, "error_message", status.error_message().c_str());
         char* jsonString = cJSON_PrintUnformatted(json);

mod_google_transcribe/google_glue_v2.cpp

@@ -349,6 +349,7 @@ static void *SWITCH_THREAD_FUNC grpc_read_thread(switch_thread_t *thread, void *
       else if (status.error_code() != 0) {
         cJSON* json = cJSON_CreateObject();
         cJSON_AddStringToObject(json, "type", "error");
+        cJSON_AddStringToObject(json, "error_cause", "stream_close");
         cJSON_AddItemToObject(json, "error_code", cJSON_CreateNumber(status.error_code()));
         cJSON_AddStringToObject(json, "error_message", status.error_message().c_str());
         char* jsonString = cJSON_PrintUnformatted(json);

mod_playht_tts/playht_glue.cpp

@@ -838,17 +838,9 @@ extern "C" {
 
     p->circularBuffer = (void *) new CircularBuffer_t(8192);
 
-    if (p->session_id) {
-      int err;
-      switch_codec_implementation_t read_impl;
-      switch_core_session_t *psession = switch_core_session_locate(p->session_id);
-      switch_core_session_get_read_impl(psession, &read_impl);
-      switch_core_session_rwunlock(psession);
-      uint32_t samples_per_second = !strcasecmp(read_impl.iananame, "g722") ? read_impl.actual_samples_per_second : read_impl.samples_per_second;
-      if (mpg123_param(mh, MPG123_FORCE_RATE, samples_per_second /*Hz*/, 0) != MPG123_OK) {
-        switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_param!\n");
-        return SWITCH_STATUS_FALSE;
-      }
+    if (mpg123_param(mh, MPG123_FORCE_RATE, p->rate /*Hz*/, 0) != MPG123_OK) {
+      switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Error mpg123_param!\n");
+      return SWITCH_STATUS_FALSE;
     }
 
     std::ostringstream api_key_stream;
@@ -881,7 +873,10 @@ extern "C" {
     curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
     //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
 
-    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+    // libcurl adding random byte to the response body that creates white noise to audio file
+    // https://github.com/curl/curl/issues/10525
+    const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
     rc = curl_multi_add_handle(global.multi, conn->easy);
     mcode_test("new_conn: curl_multi_add_handle", rc);

mod_rimelabs_tts/rimelabs_glue.cpp

@@ -394,6 +394,8 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
   std::unique_ptr<uint8_t[]> combinedData;
 
   if (conn->has_last_byte) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "rimelabs write_cb:  processing last byte from previous read\n");
+
     conn->has_last_byte = false;  // We'll handle the last_byte now, so toggle the flag off
 
     // Allocate memory for the new data array
@@ -415,6 +417,8 @@ static size_t write_cb(void *ptr, size_t size, size_t nmemb, ConnInfo_t *conn) {
 
   // If we now have an odd total, save the last byte for next time
   if ((total_bytes_to_process % sizeof(int16_t)) != 0) {
+    switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "rimelabs write_cb: got odd number of bytes %d\n", total_bytes_to_process);
+
     conn->last_byte = data[total_bytes_to_process - 1];
     conn->has_last_byte = true;
     total_bytes_to_process--;
@@ -758,6 +762,8 @@ extern "C" {
     conn->file = d->file;
     conn->body = json;
     conn->flushed = false;
+    conn->has_last_byte = false;
+    conn->last_byte = 0;
     
 
     d->circularBuffer = (void *) new CircularBuffer_t(BUFFER_GROW_SIZE);
@@ -798,7 +804,10 @@ extern "C" {
     curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
     //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
 
-    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+    // libcurl adding random byte to the response body that creates white noise to audio file
+    // https://github.com/curl/curl/issues/10525
+    const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
     rc = curl_multi_add_handle(global.multi, conn->easy);
     mcode_test("new_conn: curl_multi_add_handle", rc);

mod_whisper_tts/whisper_glue.cpp

@@ -876,7 +876,10 @@ extern "C" {
     curl_easy_setopt(easy, CURLOPT_POSTFIELDS, conn->body);
     //curl_easy_setopt(easy, CURLOPT_POSTFIELDSIZE, body.length());
 
-    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_2_0);
+    // libcurl adding random byte to the response body that creates white noise to audio file
+    // https://github.com/curl/curl/issues/10525
+    const bool disable_http_2 = switch_true(std::getenv("DISABLE_HTTP2_FOR_TTS_STREAMING"));
+    curl_easy_setopt(easy, CURLOPT_HTTP_VERSION, disable_http_2 ? CURL_HTTP_VERSION_1_1 : CURL_HTTP_VERSION_2_0);
 
     rc = curl_multi_add_handle(global.multi, conn->easy);
     mcode_test("new_conn: curl_multi_add_handle", rc);