diff --git a/esphome/components/esp_adf/speaker/esp_adf_speaker.cpp b/esphome/components/esp_adf/speaker/esp_adf_speaker.cpp
index 7cf08e1f4c..508f549abe 100644
--- a/esphome/components/esp_adf/speaker/esp_adf_speaker.cpp
+++ b/esphome/components/esp_adf/speaker/esp_adf_speaker.cpp
@@ -16,7 +16,7 @@
 namespace esphome {
 namespace esp_adf {
 
-static const size_t BUFFER_COUNT = 50;
+static const size_t BUFFER_COUNT = 20;
 
 static const char *const TAG = "esp_adf.speaker";
 
@@ -25,6 +25,8 @@ void ESPADFSpeaker::setup() {
 
   this->buffer_queue_ = xQueueCreate(BUFFER_COUNT, sizeof(DataEvent));
   this->event_queue_ = xQueueCreate(20, sizeof(TaskEvent));
+
+  ESP_LOGCONFIG(TAG, "BufferQ:%p EventQ:%p", this->buffer_queue_, this->event_queue_);
 }
 
 void ESPADFSpeaker::start() { this->state_ = speaker::STATE_STARTING; }
@@ -43,7 +45,12 @@ void ESPADFSpeaker::player_task(void *params) {
   event.type = TaskEventType::STARTING;
   xQueueSend(this_speaker->event_queue_, &event, portMAX_DELAY);
 
-  i2s_driver_config_t i2s_config = {};
+  audio_pipeline_cfg_t pipeline_cfg = {
+      .rb_size = 8 * 1024,
+  };
+  audio_pipeline_handle_t pipeline = audio_pipeline_init(&pipeline_cfg);
+
+ /* i2s_driver_config_t i2s_config = {};
   i2s_config.mode = (i2s_mode_t) (I2S_MODE_MASTER | I2S_MODE_TX);
   i2s_config.sample_rate = 16000;
   i2s_config.bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT;
@@ -59,10 +66,6 @@ void ESPADFSpeaker::player_task(void *params) {
   i2s_config.bits_per_chan = I2S_BITS_PER_CHAN_DEFAULT;
 
 
-  audio_pipeline_cfg_t pipeline_cfg = {
-      .rb_size = 8 * 1024,
-  };
-  audio_pipeline_handle_t pipeline = audio_pipeline_init(&pipeline_cfg);
 
   i2s_stream_cfg_t i2s_cfg = {};
   i2s_cfg.type = AUDIO_STREAM_WRITER;
@@ -81,8 +84,40 @@ void ESPADFSpeaker::player_task(void *params) {
   i2s_cfg.expand_src_bits = I2S_BITS_PER_SAMPLE_16BIT;
 
   audio_element_handle_t i2s_stream_writer = i2s_stream_init(&i2s_cfg);
+*/
+  i2s_stream_cfg_t i2s_cfg = {};
+  memset(&i2s_cfg, 0, sizeof(i2s_cfg));
+  i2s_cfg.type = AUDIO_STREAM_WRITER;
+  i2s_cfg.i2s_port = I2S_NUM_0;
+  i2s_cfg.use_alc = false;
+  i2s_cfg.volume = 0;
+  i2s_cfg.out_rb_size = I2S_STREAM_RINGBUFFER_SIZE;
+  i2s_cfg.task_stack = I2S_STREAM_TASK_STACK;
+  i2s_cfg.task_core = I2S_STREAM_TASK_CORE;
+  i2s_cfg.task_prio = I2S_STREAM_TASK_PRIO;
+  i2s_cfg.stack_in_ext = false;
+  i2s_cfg.multi_out_num = 0;
+  i2s_cfg.uninstall_drv = true;
+  i2s_cfg.need_expand = false;
+  i2s_cfg.expand_src_bits = I2S_BITS_PER_SAMPLE_16BIT;
+  i2s_cfg.buffer_len = I2S_STREAM_BUF_SIZE;
+  i2s_cfg.i2s_config.mode = (i2s_mode_t)(I2S_MODE_MASTER | I2S_MODE_TX);
+  i2s_cfg.i2s_config.sample_rate = 16000;
+  i2s_cfg.i2s_config.bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT;
+  i2s_cfg.i2s_config.channel_format = I2S_CHANNEL_FMT_ONLY_RIGHT;
+  i2s_cfg.i2s_config.communication_format = I2S_COMM_FORMAT_STAND_I2S;
+  i2s_cfg.i2s_config.intr_alloc_flags = ESP_INTR_FLAG_LEVEL2 | ESP_INTR_FLAG_IRAM;
+  i2s_cfg.i2s_config.dma_buf_count = 8;
+  i2s_cfg.i2s_config.dma_buf_len = 1024;
+  i2s_cfg.i2s_config.use_apll = false;
+  i2s_cfg.i2s_config.tx_desc_auto_clear = true;
+  i2s_cfg.i2s_config.fixed_mclk = 0;
+  i2s_cfg.i2s_config.bits_per_chan = I2S_BITS_PER_CHAN_DEFAULT;
 
-  rsp_filter_cfg_t rsp_cfg = {
+  audio_element_handle_t i2s_stream_writer = i2s_stream_init(&i2s_cfg);
+
+  rsp_filter_cfg_t rsp_cfg = {};
+/*
       .src_rate = 16000,
       .src_ch = 1,
       .dest_rate = 16000,
@@ -102,6 +137,25 @@ void ESPADFSpeaker::player_task(void *params) {
       .task_prio = RSP_FILTER_TASK_PRIO,
       .stack_in_ext = true,
   };
+  */
+  rsp_cfg.src_rate = 16000;
+  rsp_cfg.src_ch = 1;
+  rsp_cfg.src_bits = 16;
+  rsp_cfg.dest_rate = 16000;
+  rsp_cfg.dest_ch = 2;
+  rsp_cfg.dest_bits = 16;
+  rsp_cfg.mode = RESAMPLE_DECODE_MODE;
+  rsp_cfg.max_indata_bytes = RSP_FILTER_BUFFER_BYTE;
+  rsp_cfg.out_len_bytes = RSP_FILTER_BUFFER_BYTE;
+  rsp_cfg.type = ESP_RESAMPLE_TYPE_AUTO;
+  rsp_cfg.complexity = 2;
+  rsp_cfg.down_ch_idx = 0;
+  rsp_cfg.prefer_flag = ESP_RSP_PREFER_TYPE_SPEED;
+  rsp_cfg.out_rb_size = RSP_FILTER_RINGBUFFER_SIZE;
+  rsp_cfg.task_stack = RSP_FILTER_TASK_STACK;
+  rsp_cfg.task_core = RSP_FILTER_TASK_CORE;
+  rsp_cfg.task_prio = RSP_FILTER_TASK_PRIO;
+  rsp_cfg.stack_in_ext = true;
   audio_element_handle_t filter = rsp_filter_init(&rsp_cfg);
 
   raw_stream_cfg_t raw_cfg = {