Merge branch 'bugfix/fix_google_translate_device_example' into 'master'

bugfix(examples): Fix the google translate device example See merge request adf/esp-adf-internal!1425

Author: jason-mao

examples/cloud_services/google_translate_device/main/google_sr.c

@@ -224,6 +224,9 @@ google_sr_handle_t google_sr_init(google_sr_config_t *config)
 
  i2s_stream_cfg_t i2s_cfg = I2S_STREAM_CFG_DEFAULT();
  i2s_cfg.type = AUDIO_STREAM_READER;
+ i2s_cfg.std_cfg.slot_cfg.slot_mode = I2S_SLOT_MODE_MONO;
+ i2s_stream_set_channel_type(&i2s_cfg, I2S_CHANNEL_TYPE_ONLY_RIGHT);
+ i2s_cfg.std_cfg.clk_cfg.sample_rate_hz = config->record_sample_rates;
  sr->i2s_reader = i2s_stream_init(&i2s_cfg);
 
  http_stream_cfg_t http_cfg = {
@@ -238,7 +241,7 @@ google_sr_handle_t google_sr_init(google_sr_config_t *config)
  sr->on_begin = config->on_begin;
 
  audio_pipeline_register(sr->pipeline, sr->http_stream_writer, "sr_http");
- audio_pipeline_register(sr->pipeline, sr->i2s_reader,  "sr_i2s");
+ audio_pipeline_register(sr->pipeline, sr->i2s_reader, "sr_i2s");
  const char *link_tag[2] = {"sr_i2s", "sr_http"};
  audio_pipeline_link(sr->pipeline, &link_tag[0], 2);
  i2s_stream_set_clk(sr->i2s_reader, config->record_sample_rates, 16, 1);

examples/cloud_services/google_translate_device/main/google_tts.c

@@ -45,32 +45,30 @@
 
 static const char *TAG = "GOOGLE_TTS";
 
-
-#define GOOGLE_TTS_ENDPOINT "https://texttospeech.googleapis.com/v1beta1/text:synthesize?key=%s"
-#define GOOGLE_TTS_TEMPLATE "{"\
- "\"audioConfig\": { \"audioEncoding\" : \"MP3\", \"sampleRateHertz\": %d },"\
- "\"voice\": { \"languageCode\" : \"%s\" },"\
- "\"input\": { \"text\" : \"%s\" }"\
- "}"
-#define GOOGLE_TTS_TASK_STACK (8*1024)
+#define GOOGLE_TTS_ENDPOINT "https://texttospeech.googleapis.com/v1beta1/text:synthesize?key=%s"
+#define GOOGLE_TTS_TEMPLATE "{" \
+ "\"audioConfig\": { \"audioEncoding\" : \"MP3\", \"sampleRateHertz\": %d }," \
+ "\"voice\": { \"languageCode\" : \"%s\", \"name\" : \"%s\" }," \
+ "\"input\": { \"text\" : \"%s\" }" \
+ "}"
+#define GOOGLE_TTS_TASK_STACK (8 * 1024)
 
 typedef struct google_tts {
  audio_pipeline_handle_t pipeline;
  audio_element_handle_t i2s_writer;
  audio_element_handle_t http_stream_reader;
  audio_element_handle_t mp3_decoder;
- char *api_key;
- char *lang_code;
+ char *api_key;
+ char *lang_code;
+ char *voice_name;
  int buffer_size;
- char  *buffer;
- char  *text;
+ char *buffer;
+ char *text;
  bool is_begin;
  int tts_total_read;
  int sample_rate;
  int remain_len;
 } google_tts_t;
-
-
 static esp_err_t _http_stream_reader_event_handle(http_stream_event_msg_t *msg)
 {
  esp_http_client_handle_t http = (esp_http_client_handle_t)msg->http_client;
@@ -83,7 +81,7 @@ static esp_err_t _http_stream_reader_event_handle(http_stream_event_msg_t *msg)
  ESP_LOGI(TAG, "[ + ] HTTP client HTTP_STREAM_PRE_REQUEST, length=%d", msg->buffer_len);
  tts->tts_total_read = 0;
  tts->is_begin = true;
- int payload_len = snprintf(tts->buffer, tts->buffer_size, GOOGLE_TTS_TEMPLATE, tts->sample_rate, tts->lang_code, tts->text);
+ int payload_len = snprintf(tts->buffer, tts->buffer_size, GOOGLE_TTS_TEMPLATE, tts->sample_rate, tts->lang_code, tts->voice_name, tts->text);
  esp_http_client_set_post_field(http, tts->buffer, payload_len);
  esp_http_client_set_method(http, HTTP_METHOD_POST);
  esp_http_client_set_header(http, "Content-Type", "application/json");
@@ -121,7 +119,6 @@ static esp_err_t _http_stream_reader_event_handle(http_stream_event_msg_t *msg)
  // *data_ptr = 0;
  data_ptr++;
  process_data_len--;
-
  }
  unsigned int mp3_len = 0;
  int keep_next_time = 0;
@@ -141,11 +138,9 @@ static esp_err_t _http_stream_reader_event_handle(http_stream_event_msg_t *msg)
 
  if (msg->event_id == HTTP_STREAM_POST_REQUEST) {
  ESP_LOGI(TAG, "[ + ] HTTP client HTTP_STREAM_POST_REQUEST, write end chunked marker");
-
  }
 
  if (msg->event_id == HTTP_STREAM_FINISH_REQUEST) {
-
  }
 
  return ESP_OK;
@@ -234,8 +229,7 @@ bool google_tts_check_event_finish(google_tts_handle_t tts, audio_event_iface_ms
  return false;
 }
 
-
-esp_err_t google_tts_start(google_tts_handle_t tts, const char *text, const char *lang_code)
+esp_err_t google_tts_start(google_tts_handle_t tts, const char *text, const char *lang_code, const char *voice_name)
 {
  free(tts->lang_code);
  free(tts->text);
@@ -244,6 +238,12 @@ esp_err_t google_tts_start(google_tts_handle_t tts, const char *text, const char
  ESP_LOGE(TAG, "Error no mem");
  return ESP_ERR_NO_MEM;
  }
+ free(tts->voice_name);
+ tts->voice_name = strdup(voice_name);
+ if (tts->voice_name == NULL) {
+ ESP_LOGE(TAG, "Error no mem");
+ return ESP_ERR_NO_MEM;
+ }
  tts->text = strdup(text);
  if (tts->text == NULL) {
  free(tts->lang_code);

examples/cloud_services/google_translate_device/main/google_tts.h

@@ -53,17 +53,17 @@ typedef struct {
 google_tts_handle_t google_tts_init(google_tts_config_t *config);
 
 /**
- * @brief  Start sending text to Google Cloud Text-to-Speech and play audio received
+ * @brief Start sending text to Google Cloud Text-to-Speech and play audio received
  *
  * @param[in] tts The Text-to-Speech context
  * @param[in] text The text
  * @param[in] lang_code The language code
  *
  * @return
- * - ESP_OK
- * - ESP_FAIL
+ *  - ESP_OK
+ *  - ESP_FAIL
  */
-esp_err_t google_tts_start(google_tts_handle_t tts, const char *text, const char *lang_code);
+esp_err_t google_tts_start(google_tts_handle_t tts, const char *text, const char *lang_code, const char *voice_name);
 
 /**
  * @brief Stop playing audio from Google Cloud Text-to-Speech

examples/cloud_services/google_translate_device/main/translate_device_example.c

@@ -20,13 +20,13 @@
 #include "audio_common.h"
 #include "board.h"
 #include "esp_peripherals.h"
+#include "periph_adc_button.h"
 #include "periph_button.h"
 #include "periph_wifi.h"
 #include "periph_led.h"
 #include "google_tts.h"
 #include "google_sr.h"
 #include "google_translate.h"
-#include "board.h"
 
 #include "audio_idf_version.h"
 
@@ -38,10 +38,11 @@
 
 static const char *TAG = "GOOGLE_TRANSLATION_EXAMPLE";
 
-#define GOOGLE_SR_LANG "cmn-Hans-CN" // https://cloud.google.com/speech-to-text/docs/languages
-#define GOOGLE_TRANSLATE_LANG_FROM "zh-CN" //https://cloud.google.com/translate/docs/languages
-#define GOOGLE_TRANSLATE_LANG_TO "en" //https://cloud.google.com/translate/docs/languages
-#define GOOGLE_TTS_LANG "en-US-Wavenet-D" //https://cloud.google.com/text-to-speech/docs/voices
+#define GOOGLE_SR_LANG "cmn-Hans-CN" // https://cloud.google.com/speech-to-text/docs/languages
+#define GOOGLE_TRANSLATE_LANG_FROM "zh-CN" // https://cloud.google.com/translate/docs/languages
+#define GOOGLE_TRANSLATE_LANG_TO "en" // https://cloud.google.com/translate/docs/languages
+#define GOOGLE_TTS_LANG_CODE "en-US" // https://cloud.google.com/text-to-speech/docs/voices
+#define GOOGLE_TTS_VOICE_NAME "en-US-Wavenet-D" // https://cloud.google.com/text-to-speech/docs/voices
 
 #define EXAMPLE_RECORD_PLAYBACK_SAMPLE_RATE (16000)
 
@@ -81,10 +82,7 @@ void translate_task(void *pv)
  esp_periph_handle_t wifi_handle = periph_wifi_init(&wifi_cfg);
 
  // Initialize Button peripheral
- periph_button_cfg_t btn_cfg = {
- .gpio_mask = (1ULL << get_input_mode_id()) | (1ULL << get_input_rec_id()),
- };
- esp_periph_handle_t button_handle = periph_button_init(&btn_cfg);
+ audio_board_key_init(set);
 
  periph_led_cfg_t led_cfg = {
  .led_speed_mode = LEDC_LOW_SPEED_MODE,
@@ -94,9 +92,7 @@ void translate_task(void *pv)
  };
  led_handle = periph_led_init(&led_cfg);
 
-
  // Start wifi & button peripheral
- esp_periph_start(set, button_handle);
  esp_periph_start(set, wifi_handle);
  esp_periph_start(set, led_handle);
 
@@ -112,12 +108,14 @@ void translate_task(void *pv)
  .record_sample_rates = EXAMPLE_RECORD_PLAYBACK_SAMPLE_RATE,
  .encoding = ENCODING_LINEAR16,
  .on_begin = google_sr_begin,
+ .buffer_size = 6144,
  };
  google_sr_handle_t sr = google_sr_init(&sr_config);
 
  google_tts_config_t tts_config = {
  .api_key = CONFIG_GOOGLE_API_KEY,
  .playback_sample_rate = EXAMPLE_RECORD_PLAYBACK_SAMPLE_RATE,
+ .buffer_size = 6144,
  };
  google_tts_handle_t tts = google_tts_init(&tts_config);
 
@@ -149,7 +147,7 @@ void translate_task(void *pv)
  continue;
  }
 
- if (msg.source_type != PERIPH_ID_BUTTON) {
+ if (msg.source_type != PERIPH_ID_BUTTON && msg.source_type != PERIPH_ID_ADC_BTN) {
  continue;
  }
 
@@ -162,11 +160,12 @@ void translate_task(void *pv)
  continue;
  }
 
- if (msg.cmd == PERIPH_BUTTON_PRESSED) {
+ if (msg.cmd == PERIPH_BUTTON_PRESSED || msg.cmd == PERIPH_ADC_BUTTON_PRESSED) {
  google_tts_stop(tts);
  ESP_LOGI(TAG, "[ * ] Resuming pipeline");
  google_sr_start(sr);
- } else if (msg.cmd == PERIPH_BUTTON_RELEASE || msg.cmd == PERIPH_BUTTON_LONG_RELEASE) {
+ } else if (msg.cmd == PERIPH_BUTTON_RELEASE || msg.cmd == PERIPH_BUTTON_LONG_RELEASE \
+ || msg.cmd == PERIPH_ADC_BUTTON_RELEASE || msg.cmd == PERIPH_ADC_BUTTON_LONG_RELEASE) {
  ESP_LOGI(TAG, "[ * ] Stop pipeline");
 
  periph_led_stop(led_handle, get_green_led_gpio());
@@ -181,9 +180,8 @@ void translate_task(void *pv)
  continue;
  }
  ESP_LOGI(TAG, "Translated text = %s", translated_text);
- google_tts_start(tts, translated_text, GOOGLE_TTS_LANG);
+ google_tts_start(tts, translated_text, GOOGLE_TTS_LANG_CODE, GOOGLE_TTS_VOICE_NAME);
  }
-
  }
  ESP_LOGI(TAG, "[ 6 ] Stop audio_pipeline");
  google_sr_destroy(sr);