ESP-ADF microphone and speaker board support for s3-box

esphome/components/esp_adf/__init__.py

@@ -0,0 +1,60 @@
+import esphome.config_validation as cv
+import esphome.codegen as cg
+
+from esphome.components import esp32
+
+from esphome.const import CONF_ID
+
+# CONFLICTS_WITH = ["i2s_audio"]
+DEPENDENCIES = ["esp32"]
+
+CONF_ESP_ADF_ID = "esp_adf_id"
+
+esp_adf_ns = cg.esphome_ns.namespace("esp_adf")
+ESPADF = esp_adf_ns.class_("ESPADF", cg.Component)
+ESPADFPipeline = esp_adf_ns.class_("ESPADFPipeline", cg.Parented.template(ESPADF))
+
+SUPPORTED_BOARDS = {"esp32s3box": "CONFIG_ESP32_S3_BOX_BOARD"}
+
+
+def _validate_board(config):
+    board = esp32.get_board()
+    if board not in SUPPORTED_BOARDS:
+        raise cv.Invalid(f"Board {board} is not supported by esp-adf")
+    return config
+
+
+CONFIG_SCHEMA = cv.All(
+    cv.Schema({cv.GenerateID(): cv.declare_id(ESPADF)}),
+    _validate_board,
+    cv.only_with_esp_idf,
+)
+
+
+async def to_code(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+
+    cg.add_define("USE_ESP_ADF")
+
+    cg.add_platformio_option("build_unflags", "-Wl,--end-group")
+
+    esp32.add_idf_component(
+        name="esp-adf",
+        repo="https://github.com/espressif/esp-adf",
+        path="components",
+        ref="v2.5",
+        components=["*"],
+        submodules=["components/esp-sr", "components/esp-adf-libs"],
+    )
+
+    esp32.add_idf_component(
+        name="esp-dsp",
+        repo="https://github.com/espressif/esp-dsp",
+        ref="v1.2.0",
+    )
+
+    cg.add_platformio_option(
+        "board_build.embed_txtfiles", "components/dueros_service/duer_profile"
+    )
+    esp32.add_idf_sdkconfig_option(SUPPORTED_BOARDS[esp32.get_board()], True)

esphome/components/esp_adf/esp_adf.cpp

@@ -0,0 +1,26 @@
+#include "esp_adf.h"
+
+#ifdef USE_ESP_IDF
+
+#include <board.h>
+
+#include "esphome/core/log.h"
+
+namespace esphome {
+namespace esp_adf {
+
+static const char *const TAG = "esp_adf";
+
+void ESPADF::setup() {
+  ESP_LOGI(TAG, "Start codec chip");
+
+  audio_board_handle_t board_handle = audio_board_init();
+  audio_hal_ctrl_codec(board_handle->audio_hal, AUDIO_HAL_CODEC_MODE_BOTH, AUDIO_HAL_CTRL_START);
+}
+
+float ESPADF::get_setup_priority() const { return setup_priority::HARDWARE; }
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif

esphome/components/esp_adf/esp_adf.h

@@ -0,0 +1,31 @@
+#pragma once
+
+#ifdef USE_ESP_IDF
+
+#include "esphome/core/component.h"
+#include "esphome/core/helpers.h"
+
+namespace esphome {
+namespace esp_adf {
+class ESPADF;
+
+class ESPADFPipeline : public Parented<ESPADF> {};
+
+class ESPADF : public Component {
+ public:
+  void setup() override;
+
+  float get_setup_priority() const override;
+
+  void lock() { this->lock_.lock(); }
+  bool try_lock() { return this->lock_.try_lock(); }
+  void unlock() { this->lock_.unlock(); }
+
+ protected:
+  Mutex lock_;
+};
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif

esphome/components/esp_adf/microphone/__init__.py

@@ -0,0 +1,33 @@
+import esphome.codegen as cg
+import esphome.config_validation as cv
+from esphome.components import microphone
+from esphome.const import CONF_ID
+
+from .. import CONF_ESP_ADF_ID, ESPADF, ESPADFPipeline, esp_adf_ns
+
+AUTO_LOAD = ["esp_adf"]
+CODEOWNERS = ["@jesserockz"]
+DEPENDENCIES = ["esp32"]
+
+ESPADFMicrophone = esp_adf_ns.class_(
+    "ESPADFMicrophone", ESPADFPipeline, microphone.Microphone, cg.Component
+)
+
+
+CONFIG_SCHEMA = cv.All(
+    microphone.MICROPHONE_SCHEMA.extend(
+        {
+            cv.GenerateID(): cv.declare_id(ESPADFMicrophone),
+            cv.GenerateID(CONF_ESP_ADF_ID): cv.use_id(ESPADF),
+        }
+    ).extend(cv.COMPONENT_SCHEMA),
+    cv.only_with_esp_idf,
+)
+
+
+async def to_code(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+    await cg.register_parented(var, config[CONF_ESP_ADF_ID])
+
+    await microphone.register_microphone(var, config)

esphome/components/esp_adf/microphone/esp_adf_microphone.cpp

@@ -0,0 +1,192 @@
+#include "esp_adf_microphone.h"
+
+#ifdef USE_ESP32
+
+#include <driver/i2s.h>
+
+#include "esphome/core/hal.h"
+#include "esphome/core/log.h"
+
+#include <audio_hal.h>
+#include <filter_resample.h>
+#include <i2s_stream.h>
+#include <raw_stream.h>
+
+namespace esphome {
+namespace esp_adf {
+
+static const size_t BUFFER_SIZE = 1024;
+
+static const char *const TAG = "esp_adf.microphone";
+
+void ESPADFMicrophone::start() {
+  if (this->is_failed())
+    return;
+  if (this->state_ == microphone::STATE_STOPPING) {
+    ESP_LOGW(TAG, "Microphone is stopping, cannot start.");
+    return;
+  }
+  this->state_ = microphone::STATE_STARTING;
+}
+void ESPADFMicrophone::start_() {
+  if (!this->parent_->try_lock()) {
+    return;
+  }
+  i2s_driver_config_t i2s_config = {
+      .mode = (i2s_mode_t) (I2S_MODE_MASTER | I2S_MODE_RX),
+      .sample_rate = 44100,
+      .bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT,
+      .channel_format = I2S_CHANNEL_FMT_RIGHT_LEFT,
+      .communication_format = I2S_COMM_FORMAT_STAND_I2S,
+      .intr_alloc_flags = ESP_INTR_FLAG_LEVEL2 | ESP_INTR_FLAG_IRAM,
+      .dma_buf_count = 3,
+      .dma_buf_len = 300,
+      .use_apll = false,
+      .tx_desc_auto_clear = true,
+      .fixed_mclk = 0,
+      .mclk_multiple = I2S_MCLK_MULTIPLE_DEFAULT,
+      .bits_per_chan = I2S_BITS_PER_CHAN_DEFAULT,
+  };
+
+  ESP_LOGI(TAG, "Init pipeline");
+  audio_pipeline_cfg_t pipeline_cfg = {
+      .rb_size = 8 * 1024,
+  };
+  this->pipeline_ = audio_pipeline_init(&pipeline_cfg);
+
+  ESP_LOGI(TAG, "Init i2s stream");
+  i2s_stream_cfg_t i2s_cfg = {
+      .type = AUDIO_STREAM_READER,
+      .i2s_config = i2s_config,
+      .i2s_port = I2S_NUM_0,
+      .use_alc = false,
+      .volume = 0,
+      .out_rb_size = I2S_STREAM_RINGBUFFER_SIZE,
+      .task_stack = I2S_STREAM_TASK_STACK,
+      .task_core = I2S_STREAM_TASK_CORE,
+      .task_prio = I2S_STREAM_TASK_PRIO,
+      .stack_in_ext = false,
+      .multi_out_num = 0,
+      .uninstall_drv = true,
+      .need_expand = false,
+      .expand_src_bits = I2S_BITS_PER_SAMPLE_16BIT,
+  };
+  this->i2s_stream_reader_ = i2s_stream_init(&i2s_cfg);
+
+  ESP_LOGI(TAG, "Init filter");
+  rsp_filter_cfg_t rsp_cfg = {
+      .src_rate = 44100,
+      .src_ch = 2,
+      .dest_rate = 16000,
+      .dest_bits = 16,
+      .dest_ch = 1,
+      .src_bits = 16,
+      .mode = RESAMPLE_DECODE_MODE,
+      .max_indata_bytes = RSP_FILTER_BUFFER_BYTE,
+      .out_len_bytes = RSP_FILTER_BUFFER_BYTE,
+      .type = ESP_RESAMPLE_TYPE_AUTO,
+      .complexity = 2,
+      .down_ch_idx = 0,
+      .prefer_flag = ESP_RSP_PREFER_TYPE_SPEED,
+      .out_rb_size = RSP_FILTER_RINGBUFFER_SIZE,
+      .task_stack = RSP_FILTER_TASK_STACK,
+      .task_core = RSP_FILTER_TASK_CORE,
+      .task_prio = RSP_FILTER_TASK_PRIO,
+      .stack_in_ext = true,
+  };
+  this->filter_ = rsp_filter_init(&rsp_cfg);
+
+  ESP_LOGI(TAG, "Init raw stream");
+  raw_stream_cfg_t raw_cfg = {
+      .type = AUDIO_STREAM_READER,
+      .out_rb_size = 8 * 1024,
+  };
+  this->raw_read_ = raw_stream_init(&raw_cfg);
+
+  ESP_LOGI(TAG, "Register all elements to audio pipeline");
+  audio_pipeline_register(this->pipeline_, this->i2s_stream_reader_, "i2s");
+  audio_pipeline_register(this->pipeline_, this->filter_, "filter");
+  audio_pipeline_register(this->pipeline_, this->raw_read_, "raw");
+
+  const char *link_tag[3] = {"i2s", "filter", "raw"};
+  audio_pipeline_link(this->pipeline_, &link_tag[0], 3);
+
+  ESP_LOGI(TAG, "Starting pipeline");
+  audio_pipeline_run(this->pipeline_);
+
+  this->state_ = microphone::STATE_RUNNING;
+}
+
+void ESPADFMicrophone::stop() {
+  if (this->state_ == microphone::STATE_STOPPED || this->is_failed())
+    return;
+  if (this->state_ == microphone::STATE_STARTING) {
+    this->state_ = microphone::STATE_STOPPED;
+    return;
+  }
+  this->state_ = microphone::STATE_STOPPING;
+}
+
+void ESPADFMicrophone::stop_() {
+  ESP_LOGD(TAG, "Stopping microphone");
+  audio_pipeline_stop(this->pipeline_);
+  audio_pipeline_wait_for_stop(this->pipeline_);
+  audio_pipeline_terminate(this->pipeline_);
+
+  audio_pipeline_unregister(this->pipeline_, this->i2s_stream_reader_);
+  audio_pipeline_unregister(this->pipeline_, this->filter_);
+  audio_pipeline_unregister(this->pipeline_, this->raw_read_);
+
+  audio_pipeline_deinit(this->pipeline_);
+  audio_element_deinit(this->i2s_stream_reader_);
+  audio_element_deinit(this->filter_);
+  audio_element_deinit(this->raw_read_);
+
+  this->parent_->unlock();
+  this->state_ = microphone::STATE_STOPPED;
+  ESP_LOGD(TAG, "Microphone stopped");
+}
+
+size_t ESPADFMicrophone::read(int16_t *buf, size_t len) {
+  int bytes_read = raw_stream_read(this->raw_read_, (char *) buf, len);
+
+  if (bytes_read == ESP_FAIL) {
+    ESP_LOGW(TAG, "Error reading from I2S microphone");
+    this->status_set_warning();
+    return ESP_FAIL;
+  }
+  this->status_clear_warning();
+
+  return bytes_read;
+}
+
+void ESPADFMicrophone::read_() {
+  std::vector<int16_t> samples;
+  samples.resize(BUFFER_SIZE);
+  this->read(samples.data(), samples.size());
+
+  this->data_callbacks_.call(samples);
+}
+
+void ESPADFMicrophone::loop() {
+  switch (this->state_) {
+    case microphone::STATE_STOPPED:
+      break;
+    case microphone::STATE_STARTING:
+      this->start_();
+      break;
+    case microphone::STATE_RUNNING:
+      if (this->data_callbacks_.size() > 0) {
+        this->read_();
+      }
+      break;
+    case microphone::STATE_STOPPING:
+      this->stop_();
+      break;
+  }
+}
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif  // USE_ESP32

esphome/components/esp_adf/microphone/esp_adf_microphone.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#ifdef USE_ESP32
+
+#include "../esp_adf.h"
+
+#include "esphome/components/microphone/microphone.h"
+#include "esphome/core/component.h"
+
+#include <audio_element.h>
+#include <audio_pipeline.h>
+
+namespace esphome {
+namespace esp_adf {
+
+class ESPADFMicrophone : public ESPADFPipeline, public microphone::Microphone, public Component {
+ public:
+  void start() override;
+  void stop() override;
+
+  void loop() override;
+
+  size_t read(int16_t *buf, size_t len) override;
+
+ protected:
+  void start_();
+  void stop_();
+  void read_();
+
+  audio_pipeline_handle_t pipeline_;
+  audio_element_handle_t i2s_stream_reader_, filter_, raw_read_;
+};
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif  // USE_ESP32

esphome/components/esp_adf/speaker/__init__.py

@@ -0,0 +1,33 @@
+import esphome.codegen as cg
+import esphome.config_validation as cv
+from esphome.components import speaker
+from esphome.const import CONF_ID
+
+from .. import CONF_ESP_ADF_ID, ESPADF, ESPADFPipeline, esp_adf_ns
+
+AUTO_LOAD = ["esp_adf"]
+CODEOWNERS = ["@jesserockz"]
+DEPENDENCIES = ["esp32"]
+
+ESPADFSpeaker = esp_adf_ns.class_(
+    "ESPADFSpeaker", ESPADFPipeline, speaker.Speaker, cg.Component
+)
+
+
+CONFIG_SCHEMA = cv.All(
+    cv.Schema(
+        {
+            cv.GenerateID(): cv.declare_id(ESPADFSpeaker),
+            cv.GenerateID(CONF_ESP_ADF_ID): cv.use_id(ESPADF),
+        }
+    ).extend(cv.COMPONENT_SCHEMA),
+    cv.only_with_esp_idf,
+)
+
+
+async def to_code(config):
+    var = cg.new_Pvariable(config[CONF_ID])
+    await cg.register_component(var, config)
+    await cg.register_parented(var, config[CONF_ESP_ADF_ID])
+
+    await speaker.register_speaker(var, config)

esphome/components/esp_adf/speaker/esp_adf_speaker.cpp

@@ -0,0 +1,272 @@
+#include "esp_adf_speaker.h"
+
+#ifdef USE_ESP32
+
+#include <driver/i2s.h>
+
+#include "esphome/core/application.h"
+#include "esphome/core/hal.h"
+#include "esphome/core/log.h"
+
+#include <audio_hal.h>
+#include <filter_resample.h>
+#include <i2s_stream.h>
+#include <raw_stream.h>
+
+namespace esphome {
+namespace esp_adf {
+
+static const size_t BUFFER_COUNT = 50;
+
+static const char *const TAG = "esp_adf.speaker";
+
+void ESPADFSpeaker::setup() {
+  ESP_LOGCONFIG(TAG, "Setting up ESP ADF Speaker...");
+
+  this->buffer_queue_ = xQueueCreate(BUFFER_COUNT, sizeof(DataEvent));
+  this->event_queue_ = xQueueCreate(20, sizeof(TaskEvent));
+}
+
+void ESPADFSpeaker::start() { this->state_ = speaker::STATE_STARTING; }
+void ESPADFSpeaker::start_() {
+  if (!this->parent_->try_lock()) {
+    return;  // Waiting for another i2s component to return lock
+  }
+  this->state_ = speaker::STATE_RUNNING;
+
+  xTaskCreate(ESPADFSpeaker::player_task, "speaker_task", 8192, (void *) this, 0, &this->player_task_handle_);
+}
+
+void ESPADFSpeaker::player_task(void *params) {
+  ESPADFSpeaker *this_speaker = (ESPADFSpeaker *) params;
+
+  TaskEvent event;
+  event.type = TaskEventType::STARTING;
+  xQueueSend(this_speaker->event_queue_, &event, portMAX_DELAY);
+
+  i2s_driver_config_t i2s_config = {
+      .mode = (i2s_mode_t) (I2S_MODE_MASTER | I2S_MODE_TX),
+      .sample_rate = 16000,
+      .bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT,
+      .channel_format = I2S_CHANNEL_FMT_ONLY_RIGHT,
+      .communication_format = I2S_COMM_FORMAT_STAND_I2S,
+      .intr_alloc_flags = ESP_INTR_FLAG_LEVEL2 | ESP_INTR_FLAG_IRAM,
+      .dma_buf_count = 8,
+      .dma_buf_len = 1024,
+      .use_apll = false,
+      .tx_desc_auto_clear = true,
+      .fixed_mclk = 0,
+      .mclk_multiple = I2S_MCLK_MULTIPLE_DEFAULT,
+      .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 = {
+      .type = AUDIO_STREAM_WRITER,
+      .i2s_config = i2s_config,
+      .i2s_port = I2S_NUM_0,
+      .use_alc = false,
+      .volume = 0,
+      .out_rb_size = I2S_STREAM_RINGBUFFER_SIZE,
+      .task_stack = I2S_STREAM_TASK_STACK,
+      .task_core = I2S_STREAM_TASK_CORE,
+      .task_prio = I2S_STREAM_TASK_PRIO,
+      .stack_in_ext = false,
+      .multi_out_num = 0,
+      .uninstall_drv = true,
+      .need_expand = false,
+      .expand_src_bits = I2S_BITS_PER_SAMPLE_16BIT,
+  };
+  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,
+      .dest_bits = 16,
+      .dest_ch = 2,
+      .src_bits = 16,
+      .mode = RESAMPLE_DECODE_MODE,
+      .max_indata_bytes = RSP_FILTER_BUFFER_BYTE,
+      .out_len_bytes = RSP_FILTER_BUFFER_BYTE,
+      .type = ESP_RESAMPLE_TYPE_AUTO,
+      .complexity = 2,
+      .down_ch_idx = 0,
+      .prefer_flag = ESP_RSP_PREFER_TYPE_SPEED,
+      .out_rb_size = RSP_FILTER_RINGBUFFER_SIZE,
+      .task_stack = RSP_FILTER_TASK_STACK,
+      .task_core = RSP_FILTER_TASK_CORE,
+      .task_prio = RSP_FILTER_TASK_PRIO,
+      .stack_in_ext = true,
+  };
+  audio_element_handle_t filter = rsp_filter_init(&rsp_cfg);
+
+  raw_stream_cfg_t raw_cfg = {
+      .type = AUDIO_STREAM_WRITER,
+      .out_rb_size = 8 * 1024,
+  };
+  audio_element_handle_t raw_write = raw_stream_init(&raw_cfg);
+
+  audio_pipeline_register(pipeline, raw_write, "raw");
+  audio_pipeline_register(pipeline, filter, "filter");
+  audio_pipeline_register(pipeline, i2s_stream_writer, "i2s");
+
+  const char *link_tag[3] = {
+      "raw",
+      // "filter",
+      "i2s",
+  };
+  audio_pipeline_link(pipeline, &link_tag[0], 2);
+
+  audio_pipeline_run(pipeline);
+
+  DataEvent data_event;
+
+  event.type = TaskEventType::STARTED;
+  xQueueSend(this_speaker->event_queue_, &event, 0);
+
+  uint32_t last_received = millis();
+
+  while (true) {
+    if (xQueueReceive(this_speaker->buffer_queue_, &data_event, 0) != pdTRUE) {
+      if (millis() - last_received > 100) {
+        // No audio for 100ms, stop
+        break;
+      } else {
+        continue;
+      }
+    }
+    if (data_event.stop) {
+      // Stop signal from main thread
+      while (xQueueReceive(this_speaker->buffer_queue_, &data_event, 0) == pdTRUE) {
+        // Flush queue
+      }
+      break;
+    }
+
+    size_t remaining = data_event.len;
+    size_t current = 0;
+    if (remaining > 0)
+      last_received = millis();
+
+    while (remaining > 0) {
+      int bytes_written = raw_stream_write(raw_write, (char *) data_event.data + current, remaining);
+      if (bytes_written == ESP_FAIL) {
+        event = {.type = TaskEventType::WARNING, .err = ESP_FAIL};
+        xQueueSend(this_speaker->event_queue_, &event, 0);
+        continue;
+      }
+
+      remaining -= bytes_written;
+      current += bytes_written;
+    }
+
+    event.type = TaskEventType::PLAYING;
+    xQueueSend(this_speaker->event_queue_, &event, 0);
+  }
+
+  audio_pipeline_stop(pipeline);
+  audio_pipeline_wait_for_stop(pipeline);
+  audio_pipeline_terminate(pipeline);
+
+  event.type = TaskEventType::STOPPING;
+  xQueueSend(this_speaker->event_queue_, &event, portMAX_DELAY);
+
+  audio_pipeline_unregister(pipeline, i2s_stream_writer);
+  audio_pipeline_unregister(pipeline, filter);
+  audio_pipeline_unregister(pipeline, raw_write);
+
+  audio_pipeline_deinit(pipeline);
+  audio_element_deinit(i2s_stream_writer);
+  audio_element_deinit(filter);
+  audio_element_deinit(raw_write);
+
+  event.type = TaskEventType::STOPPED;
+  xQueueSend(this_speaker->event_queue_, &event, portMAX_DELAY);
+
+  while (true) {
+    delay(10);
+  }
+}
+
+void ESPADFSpeaker::stop() {
+  if (this->state_ == speaker::STATE_STOPPED)
+    return;
+  if (this->state_ == speaker::STATE_STARTING) {
+    this->state_ = speaker::STATE_STOPPED;
+    return;
+  }
+  this->state_ = speaker::STATE_STOPPING;
+  DataEvent data;
+  data.stop = true;
+  xQueueSendToFront(this->buffer_queue_, &data, portMAX_DELAY);
+}
+
+void ESPADFSpeaker::watch_() {
+  TaskEvent event;
+  if (xQueueReceive(this->event_queue_, &event, 0) == pdTRUE) {
+    switch (event.type) {
+      case TaskEventType::STARTING:
+      case TaskEventType::STARTED:
+      case TaskEventType::STOPPING:
+        break;
+      case TaskEventType::PLAYING:
+        this->status_clear_warning();
+        break;
+      case TaskEventType::STOPPED:
+        this->parent_->unlock();
+        this->state_ = speaker::STATE_STOPPED;
+        vTaskDelete(this->player_task_handle_);
+        this->player_task_handle_ = nullptr;
+        break;
+      case TaskEventType::WARNING:
+        ESP_LOGW(TAG, "Error writing to pipeline: %s", esp_err_to_name(event.err));
+        this->status_set_warning();
+        break;
+    }
+  }
+}
+
+void ESPADFSpeaker::loop() {
+  switch (this->state_) {
+    case speaker::STATE_STARTING:
+      this->start_();
+      break;
+    case speaker::STATE_RUNNING:
+      this->watch_();
+      break;
+    case speaker::STATE_STOPPING:
+    case speaker::STATE_STOPPED:
+      break;
+  }
+}
+
+size_t ESPADFSpeaker::play(const uint8_t *data, size_t length) {
+  if (this->state_ != speaker::STATE_RUNNING && this->state_ != speaker::STATE_STARTING) {
+    this->start();
+  }
+  size_t remaining = length;
+  size_t index = 0;
+  while (remaining > 0) {
+    DataEvent event;
+    event.stop = false;
+    size_t to_send_length = std::min(remaining, BUFFER_SIZE);
+    event.len = to_send_length;
+    memcpy(event.data, data + index, to_send_length);
+    if (xQueueSend(this->buffer_queue_, &event, 0) != pdTRUE) {
+      return index;  // Queue full
+    }
+    remaining -= to_send_length;
+    index += to_send_length;
+  }
+  return index;
+}
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif  // USE_ESP32

esphome/components/esp_adf/speaker/esp_adf_speaker.h

@@ -0,0 +1,68 @@
+#pragma once
+
+#ifdef USE_ESP32
+
+#include "../esp_adf.h"
+
+#include <freertos/FreeRTOS.h>
+#include <freertos/queue.h>
+
+#include "esphome/components/speaker/speaker.h"
+#include "esphome/core/component.h"
+#include "esphome/core/helpers.h"
+
+#include <audio_element.h>
+#include <audio_pipeline.h>
+
+namespace esphome {
+namespace esp_adf {
+
+static const size_t BUFFER_SIZE = 1024;
+
+enum class TaskEventType : uint8_t {
+  STARTING = 0,
+  STARTED,
+  PLAYING,
+  STOPPING,
+  STOPPED,
+  WARNING = 255,
+};
+
+struct TaskEvent {
+  TaskEventType type;
+  esp_err_t err;
+};
+
+struct DataEvent {
+  bool stop;
+  size_t len;
+  uint8_t data[BUFFER_SIZE];
+};
+
+class ESPADFSpeaker : public ESPADFPipeline, public speaker::Speaker, public Component {
+ public:
+  float get_setup_priority() const override { return esphome::setup_priority::LATE; }
+
+  void setup() override;
+  void loop() override;
+
+  void start() override;
+  void stop() override;
+
+  size_t play(const uint8_t *data, size_t length) override;
+
+ protected:
+  void start_();
+  void watch_();
+
+  static void player_task(void *params);
+
+  TaskHandle_t player_task_handle_{nullptr};
+  QueueHandle_t buffer_queue_;
+  QueueHandle_t event_queue_;
+};
+
+}  // namespace esp_adf
+}  // namespace esphome
+
+#endif  // USE_ESP32

esphome/core/defines.h

@@ -62,6 +62,7 @@
 // IDF-specific feature flags
 #ifdef USE_ESP_IDF
 #define USE_MQTT_IDF_ENQUEUE
+#define USE_ESP_ADF
 #endif
 
 // ESP32-specific feature flags