Merge branch 'esphome:dev' into nvds-i2s_lock-241024

.github/actions/restore-python/action.yml

@@ -22,7 +22,7 @@ runs:
         python-version: ${{ inputs.python-version }}
     - name: Restore Python virtual environment
       id: cache-venv
-      uses: actions/cache/restore@v4.1.1
+      uses: actions/cache/restore@v4.1.2
       with:
         path: venv
         # yamllint disable-line rule:line-length

.github/workflows/ci.yml

@@ -46,7 +46,7 @@ jobs:
           python-version: ${{ env.DEFAULT_PYTHON }}
       - name: Restore Python virtual environment
         id: cache-venv
-        uses: actions/cache@v4.1.1
+        uses: actions/cache@v4.1.2
         with:
           path: venv
           # yamllint disable-line rule:line-length
@@ -302,14 +302,14 @@ jobs:
 
       - name: Cache platformio
         if: github.ref == 'refs/heads/dev'
-        uses: actions/cache@v4.1.1
+        uses: actions/cache@v4.1.2
         with:
           path: ~/.platformio
           key: platformio-${{ matrix.pio_cache_key }}
 
       - name: Cache platformio
         if: github.ref != 'refs/heads/dev'
-        uses: actions/cache/restore@v4.1.1
+        uses: actions/cache/restore@v4.1.2
         with:
           path: ~/.platformio
           key: platformio-${{ matrix.pio_cache_key }}

CODEOWNERS

@@ -202,7 +202,7 @@ esphome/components/i2c_device/* @gabest11
 esphome/components/i2s_audio/* @jesserockz
 esphome/components/i2s_audio/media_player/* @jesserockz
 esphome/components/i2s_audio/microphone/* @jesserockz
-esphome/components/i2s_audio/speaker/* @jesserockz
+esphome/components/i2s_audio/speaker/* @jesserockz @kahrendt
 esphome/components/iaqcore/* @yozik04
 esphome/components/ili9xxx/* @clydebarrow @nielsnl68
 esphome/components/improv_base/* @esphome/core
@@ -377,7 +377,7 @@ esphome/components/smt100/* @piechade
 esphome/components/sn74hc165/* @jesserockz
 esphome/components/socket/* @esphome/core
 esphome/components/sonoff_d1/* @anatoly-savchenkov
-esphome/components/speaker/* @jesserockz
+esphome/components/speaker/* @jesserockz @kahrendt
 esphome/components/spi/* @clydebarrow @esphome/core
 esphome/components/spi_device/* @clydebarrow
 esphome/components/spi_led_strip/* @clydebarrow

esphome/components/esp32/boards.py

@@ -103,6 +103,173 @@ ESP32_BOARD_PINS = {
         "LED": 13,
         "LED_BUILTIN": 13,
     },
+    "adafruit_feather_esp32s3": {
+        "BUTTON": 0,
+        "A0": 18,
+        "A1": 17,
+        "A2": 16,
+        "A3": 15,
+        "A4": 14,
+        "A5": 8,
+        "SCK": 36,
+        "MOSI": 35,
+        "MISO": 37,
+        "RX": 38,
+        "TX": 39,
+        "SCL": 4,
+        "SDA": 3,
+        "NEOPIXEL": 33,
+        "PIN_NEOPIXEL": 33,
+        "NEOPIXEL_POWER": 21,
+        "I2C_POWER": 7,
+        "LED": 13,
+        "LED_BUILTIN": 13,
+    },
+    "adafruit_feather_esp32s3_nopsram": {
+        "BUTTON": 0,
+        "A0": 18,
+        "A1": 17,
+        "A2": 16,
+        "A3": 15,
+        "A4": 14,
+        "A5": 8,
+        "SCK": 36,
+        "MOSI": 35,
+        "MISO": 37,
+        "RX": 38,
+        "TX": 39,
+        "SCL": 4,
+        "SDA": 3,
+        "NEOPIXEL": 33,
+        "PIN_NEOPIXEL": 33,
+        "NEOPIXEL_POWER": 21,
+        "I2C_POWER": 7,
+        "LED": 13,
+        "LED_BUILTIN": 13,
+    },
+    "adafruit_feather_esp32s3_tft": {
+        "BUTTON": 0,
+        "A0": 18,
+        "A1": 17,
+        "A2": 16,
+        "A3": 15,
+        "A4": 14,
+        "A5": 8,
+        "SCK": 36,
+        "MOSI": 35,
+        "MISO": 37,
+        "RX": 2,
+        "TX": 1,
+        "SCL": 41,
+        "SDA": 42,
+        "NEOPIXEL": 33,
+        "PIN_NEOPIXEL": 33,
+        "NEOPIXEL_POWER": 34,
+        "TFT_I2C_POWER": 21,
+        "TFT_CS": 7,
+        "TFT_DC": 39,
+        "TFT_RESET": 40,
+        "TFT_BACKLIGHT": 45,
+        "LED": 13,
+        "LED_BUILTIN": 13,
+    },
+    "adafruit_funhouse_esp32s2": {
+        "BUTTON_UP": 5,
+        "BUTTON_DOWN": 3,
+        "BUTTON_SELECT": 4,
+        "DOTSTAR_DATA": 14,
+        "DOTSTAR_CLOCK": 15,
+        "PIR_SENSE": 16,
+        "A0": 17,
+        "A1": 2,
+        "A2": 1,
+        "CAP6": 6,
+        "CAP7": 7,
+        "CAP8": 8,
+        "CAP9": 9,
+        "CAP10": 10,
+        "CAP11": 11,
+        "CAP12": 12,
+        "CAP13": 13,
+        "SPEAKER": 42,
+        "LED": 37,
+        "LIGHT": 18,
+        "TFT_MOSI": 35,
+        "TFT_SCK": 36,
+        "TFT_CS": 40,
+        "TFT_DC": 39,
+        "TFT_RESET": 41,
+        "TFT_BACKLIGHT": 21,
+        "RED_LED": 31,
+        "BUTTON": 0,
+    },
+    "adafruit_itsybitsy_esp32": {
+        "A0": 25,
+        "A1": 26,
+        "A2": 4,
+        "A3": 38,
+        "A4": 37,
+        "A5": 36,
+        "SCK": 19,
+        "MOSI": 21,
+        "MISO": 22,
+        "SCL": 27,
+        "SDA": 15,
+        "TX": 20,
+        "RX": 8,
+        "NEOPIXEL": 0,
+        "PIN_NEOPIXEL": 0,
+        "NEOPIXEL_POWER": 2,
+        "BUTTON": 35,
+    },
+    "adafruit_magtag29_esp32s2": {
+        "A1": 18,
+        "BUTTON_A": 15,
+        "BUTTON_B": 14,
+        "BUTTON_C": 12,
+        "BUTTON_D": 11,
+        "SDA": 33,
+        "SCL": 34,
+        "SPEAKER": 17,
+        "SPEAKER_ENABLE": 16,
+        "VOLTAGE_MONITOR": 4,
+        "ACCELEROMETER_INT": 9,
+        "ACCELEROMETER_INTERRUPT": 9,
+        "LIGHT": 3,
+        "NEOPIXEL": 1,
+        "PIN_NEOPIXEL": 1,
+        "NEOPIXEL_POWER": 21,
+        "EPD_BUSY": 5,
+        "EPD_RESET": 6,
+        "EPD_DC": 7,
+        "EPD_CS": 8,
+        "EPD_MOSI": 35,
+        "EPD_SCK": 36,
+        "EPD_MISO": 37,
+        "BUTTON": 0,
+        "LED": 13,
+        "LED_BUILTIN": 13,
+    },
+    "adafruit_metro_esp32s2": {
+        "A0": 17,
+        "A1": 18,
+        "A2": 1,
+        "A3": 2,
+        "A4": 3,
+        "A5": 4,
+        "RX": 38,
+        "TX": 37,
+        "SCL": 34,
+        "SDA": 33,
+        "MISO": 37,
+        "SCK": 36,
+        "MOSI": 35,
+        "NEOPIXEL": 45,
+        "PIN_NEOPIXEL": 45,
+        "LED": 42,
+        "LED_BUILTIN": 42,
+        "BUTTON": 0,
+    },
     "adafruit_qtpy_esp32c3": {
         "A0": 4,
         "A1": 3,
@@ -141,6 +308,26 @@ ESP32_BOARD_PINS = {
         "BUTTON": 0,
         "SWITCH": 0,
     },
+    "adafruit_qtpy_esp32s3_nopsram": {
+        "A0": 18,
+        "A1": 17,
+        "A2": 9,
+        "A3": 8,
+        "SDA": 7,
+        "SCL": 6,
+        "MOSI": 35,
+        "MISO": 37,
+        "SCK": 36,
+        "RX": 16,
+        "TX": 5,
+        "SDA1": 41,
+        "SCL1": 40,
+        "NEOPIXEL": 39,
+        "PIN_NEOPIXEL": 39,
+        "NEOPIXEL_POWER": 38,
+        "BUTTON": 0,
+        "SWITCH": 0,
+    },
     "adafruit_qtpy_esp32": {
         "A0": 26,
         "A1": 25,
@@ -1068,7 +1255,18 @@ ESP32_BOARD_PINS = {
         "_VBAT": 35,
     },
     "wemosbat": {"LED": 16},
-    "wesp32": {"MISO": 32, "SCL": 4, "SDA": 15},
+    "wesp32": {
+        "MISO": 32,
+        "MOSI": 23,
+        "SCK": 18,
+        "SCL": 4,
+        "SDA": 15,
+        "MISO1": 12,
+        "MOSI1": 13,
+        "SCK1": 14,
+        "SCL1": 5,
+        "SDA1": 33,
+    },
     "widora-air": {
         "A1": 39,
         "A2": 35,

esphome/components/i2s_audio/speaker/__init__.py

@@ -17,7 +17,7 @@ from .. import (
 )
 
 AUTO_LOAD = ["audio"]
-CODEOWNERS = ["@jesserockz"]
+CODEOWNERS = ["@jesserockz", "@kahrendt"]
 DEPENDENCIES = ["i2s_audio"]
 
 I2SAudioSpeaker = i2s_audio_ns.class_(

esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp

@@ -32,6 +32,7 @@ enum SpeakerEventGroupBits : uint32_t {
   STATE_RUNNING = (1 << 11),
   STATE_STOPPING = (1 << 12),
   STATE_STOPPED = (1 << 13),
+  ERR_INVALID_FORMAT = (1 << 14),
   ERR_TASK_FAILED_TO_START = (1 << 15),
   ERR_ESP_INVALID_STATE = (1 << 16),
   ERR_ESP_INVALID_ARG = (1 << 17),
@@ -104,16 +105,6 @@ void I2SAudioSpeaker::setup() {
 void I2SAudioSpeaker::loop() {
   uint32_t event_group_bits = xEventGroupGetBits(this->event_group_);
 
-  if (event_group_bits & SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START) {
-    this->status_set_error("Failed to start speaker task");
-  }
-
-  if (event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS) {
-    uint32_t error_bits = event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS;
-    ESP_LOGW(TAG, "Error writing to I2S: %s", esp_err_to_name(err_bit_to_esp_err(error_bits)));
-    this->status_set_warning();
-  }
-
   if (event_group_bits & SpeakerEventGroupBits::STATE_STARTING) {
     ESP_LOGD(TAG, "Starting Speaker");
     this->state_ = speaker::STATE_STARTING;
@@ -139,12 +130,64 @@ void I2SAudioSpeaker::loop() {
       this->speaker_task_handle_ = nullptr;
     }
   }
+
+  if (event_group_bits & SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START) {
+    this->status_set_error("Failed to start speaker task");
+    xEventGroupClearBits(this->event_group_, SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START);
+  }
+
+  if (event_group_bits & SpeakerEventGroupBits::ERR_INVALID_FORMAT) {
+    this->status_set_error("Failed to adjust I2S bus to match the incoming audio");
+    ESP_LOGE(TAG,
+             "Incompatible audio format: sample rate = %" PRIu32 ", channels = %" PRIu8 ", bits per sample = %" PRIu8,
+             this->audio_stream_info_.sample_rate, this->audio_stream_info_.channels,
+             this->audio_stream_info_.bits_per_sample);
+  }
+
+  if (event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS) {
+    uint32_t error_bits = event_group_bits & SpeakerEventGroupBits::ALL_ERR_ESP_BITS;
+    ESP_LOGW(TAG, "Error writing to I2S: %s", esp_err_to_name(err_bit_to_esp_err(error_bits)));
+    this->status_set_warning();
+  }
 }
 
 void I2SAudioSpeaker::set_volume(float volume) {
   this->volume_ = volume;
-  ssize_t decibel_index = remap<ssize_t, float>(volume, 0.0f, 1.0f, 0, Q15_VOLUME_SCALING_FACTORS.size() - 1);
-  this->q15_volume_factor_ = Q15_VOLUME_SCALING_FACTORS[decibel_index];
+#ifdef USE_AUDIO_DAC
+  if (this->audio_dac_ != nullptr) {
+    if (volume > 0.0) {
+      this->audio_dac_->set_mute_off();
+    }
+    this->audio_dac_->set_volume(volume);
+  } else
+#endif
+  {
+    // Fallback to software volume control by using a Q15 fixed point scaling factor
+    ssize_t decibel_index = remap<ssize_t, float>(volume, 0.0f, 1.0f, 0, Q15_VOLUME_SCALING_FACTORS.size() - 1);
+    this->q15_volume_factor_ = Q15_VOLUME_SCALING_FACTORS[decibel_index];
+  }
+}
+
+void I2SAudioSpeaker::set_mute_state(bool mute_state) {
+  this->mute_state_ = mute_state;
+#ifdef USE_AUDIO_DAC
+  if (this->audio_dac_) {
+    if (mute_state) {
+      this->audio_dac_->set_mute_on();
+    } else {
+      this->audio_dac_->set_mute_off();
+    }
+  } else
+#endif
+  {
+    if (mute_state) {
+      // Fallback to software volume control and scale by 0
+      this->q15_volume_factor_ = 0;
+    } else {
+      // Revert to previous volume when unmuting
+      this->set_volume(this->volume_);
+    }
+  }
 }
 
 size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length, TickType_t ticks_to_wait) {
@@ -275,6 +318,9 @@ void I2SAudioSpeaker::speaker_task(void *params) {
         i2s_zero_dma_buffer(this_speaker->parent_->get_port());
       }
     }
+  } else {
+    // Couldn't configure the I2S port to be compatible with the incoming audio
+    xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_INVALID_FORMAT);
   }
   i2s_zero_dma_buffer(this_speaker->parent_->get_port());
 
@@ -288,7 +334,7 @@ void I2SAudioSpeaker::speaker_task(void *params) {
 }
 
 void I2SAudioSpeaker::start() {
-  if (this->is_failed())
+  if (this->is_failed() || this->status_has_error())
     return;
   if ((this->state_ == speaker::STATE_STARTING) || (this->state_ == speaker::STATE_RUNNING))
     return;

esphome/components/i2s_audio/speaker/i2s_audio_speaker.h

@@ -49,11 +49,17 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp
 
   bool has_buffered_data() const override;
 
-  /// @brief Sets the volume of the speaker. It is implemented as a software volume control.
-  /// Overrides the default setter to convert the floating point volume to a Q15 fixed-point factor.
-  /// @param volume
+  /// @brief Sets the volume of the speaker. Uses the speaker's configured audio dac component. If unavailble, it is
+  /// implemented as a software volume control. Overrides the default setter to convert the floating point volume to a
+  /// Q15 fixed-point factor.
+  /// @param volume between 0.0 and 1.0
   void set_volume(float volume) override;
-  float get_volume() override { return this->volume_; }
+
+  /// @brief Mutes or unmute the speaker. Uses the speaker's configured audio dac component. If unavailble, it is
+  /// implemented as a software volume control. Overrides the default setter to convert the floating point volume to a
+  /// Q15 fixed-point factor.
+  /// @param mute_state true for muting, false for unmuting
+  void set_mute_state(bool mute_state) override;
 
  protected:
   /// @brief Function for the FreeRTOS task handling audio output.

esphome/components/rp2040/__init__.py

@@ -17,7 +17,7 @@ from esphome.const import (
     PLATFORM_RP2040,
 )
 from esphome.core import CORE, EsphomeError, coroutine_with_priority
-from esphome.helpers import copy_file_if_changed, mkdir_p, write_file
+from esphome.helpers import copy_file_if_changed, mkdir_p, write_file, read_file
 
 from .const import KEY_BOARD, KEY_PIO_FILES, KEY_RP2040, rp2040_ns
 
@@ -230,11 +230,14 @@ def generate_pio_files() -> bool:
 
 
 # Called by writer.py
-def copy_files() -> bool:
+def copy_files():
     dir = os.path.dirname(__file__)
     post_build_file = os.path.join(dir, "post_build.py.script")
     copy_file_if_changed(
         post_build_file,
         CORE.relative_build_path("post_build.py"),
     )
-    return generate_pio_files()
+    if generate_pio_files():
+        path = CORE.relative_src_path("esphome.h")
+        content = read_file(path).rstrip("\n")
+        write_file(path, content + '\n#include "pio_includes.h"\n')

esphome/components/speaker/__init__.py

@@ -1,15 +1,18 @@
 from esphome import automation
 from esphome.automation import maybe_simple_id
 import esphome.codegen as cg
+from esphome.components import audio_dac
 import esphome.config_validation as cv
 from esphome.const import CONF_DATA, CONF_ID, CONF_VOLUME
 from esphome.core import CORE
 from esphome.coroutine import coroutine_with_priority
 
-CODEOWNERS = ["@jesserockz"]
+CODEOWNERS = ["@jesserockz", "@kahrendt"]
 
 IS_PLATFORM_COMPONENT = True
 
+CONF_AUDIO_DAC = "audio_dac"
+
 speaker_ns = cg.esphome_ns.namespace("speaker")
 
 Speaker = speaker_ns.class_("Speaker")
@@ -26,6 +29,12 @@ FinishAction = speaker_ns.class_(
 VolumeSetAction = speaker_ns.class_(
     "VolumeSetAction", automation.Action, cg.Parented.template(Speaker)
 )
+MuteOnAction = speaker_ns.class_(
+    "MuteOnAction", automation.Action, cg.Parented.template(Speaker)
+)
+MuteOffAction = speaker_ns.class_(
+    "MuteOffAction", automation.Action, cg.Parented.template(Speaker)
+)
 
 
 IsPlayingCondition = speaker_ns.class_("IsPlayingCondition", automation.Condition)
@@ -33,7 +42,9 @@ IsStoppedCondition = speaker_ns.class_("IsStoppedCondition", automation.Conditio
 
 
 async def setup_speaker_core_(var, config):
-    pass
+    if audio_dac_config := config.get(CONF_AUDIO_DAC):
+        aud_dac = await cg.get_variable(audio_dac_config)
+        cg.add(var.set_audio_dac(aud_dac))
 
 
 async def register_speaker(var, config):
@@ -42,8 +53,11 @@ async def register_speaker(var, config):
     await setup_speaker_core_(var, config)
 
 
-SPEAKER_SCHEMA = cv.Schema({})
-
+SPEAKER_SCHEMA = cv.Schema(
+    {
+        cv.Optional(CONF_AUDIO_DAC): cv.use_id(audio_dac.AudioDac),
+    }
+)
 
 SPEAKER_AUTOMATION_SCHEMA = maybe_simple_id({cv.GenerateID(): cv.use_id(Speaker)})
 
@@ -113,6 +127,15 @@ async def speaker_volume_set_action(config, action_id, template_arg, args):
     return var
 
 
+@automation.register_action(
+    "speaker.mute_off", MuteOffAction, SPEAKER_AUTOMATION_SCHEMA
+)
+@automation.register_action("speaker.mute_on", MuteOnAction, SPEAKER_AUTOMATION_SCHEMA)
+async def speaker_mute_action_to_code(config, action_id, template_arg, args):
+    paren = await cg.get_variable(config[CONF_ID])
+    return cg.new_Pvariable(action_id, template_arg, paren)
+
+
 @coroutine_with_priority(100.0)
 async def to_code(config):
     cg.add_global(speaker_ns.using)

esphome/components/speaker/automation.h

@@ -39,6 +39,26 @@ template<typename... Ts> class VolumeSetAction : public Action<Ts...>, public Pa
   void play(Ts... x) override { this->parent_->set_volume(this->volume_.value(x...)); }
 };
 
+template<typename... Ts> class MuteOnAction : public Action<Ts...> {
+ public:
+  explicit MuteOnAction(Speaker *speaker) : speaker_(speaker) {}
+
+  void play(Ts... x) override { this->speaker_->set_mute_state(true); }
+
+ protected:
+  Speaker *speaker_;
+};
+
+template<typename... Ts> class MuteOffAction : public Action<Ts...> {
+ public:
+  explicit MuteOffAction(Speaker *speaker) : speaker_(speaker) {}
+
+  void play(Ts... x) override { this->speaker_->set_mute_state(false); }
+
+ protected:
+  Speaker *speaker_;
+};
+
 template<typename... Ts> class StopAction : public Action<Ts...>, public Parented<Speaker> {
  public:
   void play(Ts... x) override { this->parent_->stop(); }

esphome/components/speaker/speaker.h

@@ -8,7 +8,12 @@
 #include <freertos/FreeRTOS.h>
 #endif
 
+#include "esphome/core/defines.h"
+
 #include "esphome/components/audio/audio.h"
+#ifdef USE_AUDIO_DAC
+#include "esphome/components/audio_dac/audio_dac.h"
+#endif
 
 namespace esphome {
 namespace speaker {
@@ -56,9 +61,35 @@ class Speaker {
   bool is_running() const { return this->state_ == STATE_RUNNING; }
   bool is_stopped() const { return this->state_ == STATE_STOPPED; }
 
-  // Volume control must be implemented by each speaker component, otherwise it will have no effect.
-  virtual void set_volume(float volume) { this->volume_ = volume; };
-  virtual float get_volume() { return this->volume_; }
+  // Volume control is handled by a configured audio dac component. Individual speaker components can
+  // override and implement in software if an audio dac isn't available.
+  virtual void set_volume(float volume) {
+    this->volume_ = volume;
+#ifdef USE_AUDIO_DAC
+    if (this->audio_dac_ != nullptr) {
+      this->audio_dac_->set_volume(volume);
+    }
+#endif
+  };
+  float get_volume() { return this->volume_; }
+
+  virtual void set_mute_state(bool mute_state) {
+    this->mute_state_ = mute_state;
+#ifdef USE_AUDIO_DAC
+    if (this->audio_dac_) {
+      if (mute_state) {
+        this->audio_dac_->set_mute_on();
+      } else {
+        this->audio_dac_->set_mute_off();
+      }
+    }
+#endif
+  }
+  bool get_mute_state() { return this->mute_state_; }
+
+#ifdef USE_AUDIO_DAC
+  void set_audio_dac(audio_dac::AudioDac *audio_dac) { this->audio_dac_ = audio_dac; }
+#endif
 
   void set_audio_stream_info(const audio::AudioStreamInfo &audio_stream_info) {
     this->audio_stream_info_ = audio_stream_info;
@@ -68,6 +99,11 @@ class Speaker {
   State state_{STATE_STOPPED};
   audio::AudioStreamInfo audio_stream_info_;
   float volume_{1.0f};
+  bool mute_state_{false};
+
+#ifdef USE_AUDIO_DAC
+  audio_dac::AudioDac *audio_dac_{nullptr};
+#endif
 };
 
 }  // namespace speaker

esphome/components/voice_assistant/voice_assistant.cpp

@@ -433,16 +433,18 @@ void VoiceAssistant::loop() {
 
 #ifdef USE_SPEAKER
 void VoiceAssistant::write_speaker_() {
-  if (this->speaker_buffer_size_ > 0) {
-    size_t write_chunk = std::min<size_t>(this->speaker_buffer_size_, 4 * 1024);
-    size_t written = this->speaker_->play(this->speaker_buffer_, write_chunk);
-    if (written > 0) {
-      memmove(this->speaker_buffer_, this->speaker_buffer_ + written, this->speaker_buffer_size_ - written);
-      this->speaker_buffer_size_ -= written;
-      this->speaker_buffer_index_ -= written;
-      this->set_timeout("speaker-timeout", 5000, [this]() { this->speaker_->stop(); });
-    } else {
-      ESP_LOGV(TAG, "Speaker buffer full, trying again next loop");
+  if ((this->speaker_ != nullptr) && (this->speaker_buffer_ != nullptr)) {
+    if (this->speaker_buffer_size_ > 0) {
+      size_t write_chunk = std::min<size_t>(this->speaker_buffer_size_, 4 * 1024);
+      size_t written = this->speaker_->play(this->speaker_buffer_, write_chunk);
+      if (written > 0) {
+        memmove(this->speaker_buffer_, this->speaker_buffer_ + written, this->speaker_buffer_size_ - written);
+        this->speaker_buffer_size_ -= written;
+        this->speaker_buffer_index_ -= written;
+        this->set_timeout("speaker-timeout", 5000, [this]() { this->speaker_->stop(); });
+      } else {
+        ESP_LOGV(TAG, "Speaker buffer full, trying again next loop");
+      }
     }
   }
 }
@@ -772,16 +774,20 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
     }
     case api::enums::VOICE_ASSISTANT_TTS_STREAM_START: {
 #ifdef USE_SPEAKER
-      this->wait_for_stream_end_ = true;
-      ESP_LOGD(TAG, "TTS stream start");
-      this->defer([this] { this->tts_stream_start_trigger_->trigger(); });
+      if (this->speaker_ != nullptr) {
+        this->wait_for_stream_end_ = true;
+        ESP_LOGD(TAG, "TTS stream start");
+        this->defer([this] { this->tts_stream_start_trigger_->trigger(); });
+      }
 #endif
       break;
     }
     case api::enums::VOICE_ASSISTANT_TTS_STREAM_END: {
 #ifdef USE_SPEAKER
-      this->stream_ended_ = true;
-      ESP_LOGD(TAG, "TTS stream end");
+      if (this->speaker_ != nullptr) {
+        this->stream_ended_ = true;
+        ESP_LOGD(TAG, "TTS stream end");
+      }
 #endif
       break;
     }
@@ -802,14 +808,16 @@ void VoiceAssistant::on_event(const api::VoiceAssistantEventResponse &msg) {
 
 void VoiceAssistant::on_audio(const api::VoiceAssistantAudio &msg) {
 #ifdef USE_SPEAKER  // We should never get to this function if there is no speaker anyway
-  if (this->speaker_buffer_index_ + msg.data.length() < SPEAKER_BUFFER_SIZE) {
-    memcpy(this->speaker_buffer_ + this->speaker_buffer_index_, msg.data.data(), msg.data.length());
-    this->speaker_buffer_index_ += msg.data.length();
-    this->speaker_buffer_size_ += msg.data.length();
-    this->speaker_bytes_received_ += msg.data.length();
-    ESP_LOGV(TAG, "Received audio: %u bytes from API", msg.data.length());
-  } else {
-    ESP_LOGE(TAG, "Cannot receive audio, buffer is full");
+  if ((this->speaker_ != nullptr) && (this->speaker_buffer_ != nullptr)) {
+    if (this->speaker_buffer_index_ + msg.data.length() < SPEAKER_BUFFER_SIZE) {
+      memcpy(this->speaker_buffer_ + this->speaker_buffer_index_, msg.data.data(), msg.data.length());
+      this->speaker_buffer_index_ += msg.data.length();
+      this->speaker_buffer_size_ += msg.data.length();
+      this->speaker_bytes_received_ += msg.data.length();
+      ESP_LOGV(TAG, "Received audio: %u bytes from API", msg.data.length());
+    } else {
+      ESP_LOGE(TAG, "Cannot receive audio, buffer is full");
+    }
   }
 #endif
 }

esphome/components/voice_assistant/voice_assistant.h

@@ -250,7 +250,7 @@ class VoiceAssistant : public Component {
 #ifdef USE_SPEAKER
   void write_speaker_();
   speaker::Speaker *speaker_{nullptr};
-  uint8_t *speaker_buffer_;
+  uint8_t *speaker_buffer_{nullptr};
   size_t speaker_buffer_index_{0};
   size_t speaker_buffer_size_{0};
   size_t speaker_bytes_received_{0};
@@ -282,8 +282,8 @@ class VoiceAssistant : public Component {
   float volume_multiplier_;
   uint32_t conversation_timeout_;
 
-  uint8_t *send_buffer_;
-  int16_t *input_buffer_;
+  uint8_t *send_buffer_{nullptr};
+  int16_t *input_buffer_{nullptr};
 
   bool continuous_{false};
   bool silence_detection_;

esphome/writer.py

@@ -1,3 +1,4 @@
+import importlib
 import logging
 import os
 from pathlib import Path
@@ -299,25 +300,13 @@ def copy_src_tree():
         CORE.relative_src_path("esphome", "core", "version.h"), generate_version_h()
     )
 
-    if CORE.is_esp32:
-        from esphome.components.esp32 import copy_files
-
+    platform = "esphome.components." + CORE.target_platform
+    try:
+        module = importlib.import_module(platform)
+        copy_files = getattr(module, "copy_files")
         copy_files()
-
-    elif CORE.is_esp8266:
-        from esphome.components.esp8266 import copy_files
-
-        copy_files()
-
-    elif CORE.is_rp2040:
-        from esphome.components.rp2040 import copy_files
-
-        (pio) = copy_files()
-        if pio:
-            write_file_if_changed(
-                CORE.relative_src_path("esphome.h"),
-                ESPHOME_H_FORMAT.format(include_s + '\n#include "pio_includes.h"'),
-            )
+    except AttributeError:
+        pass
 
 
 def generate_defines_h():

tests/components/speaker/test.esp32-ard.yaml

@@ -1,6 +1,8 @@
 esphome:
   on_boot:
     then:
+      - speaker.mute_on:
+      - speaker.mute_off:
       - if:
           condition: speaker.is_stopped
           then:

tests/components/speaker/test.esp32-c3-ard.yaml

@@ -1,6 +1,8 @@
 esphome:
   on_boot:
     then:
+      - speaker.mute_on:
+      - speaker.mute_off:
       - if:
           condition: speaker.is_stopped
           then:

tests/components/speaker/test.esp32-c3-idf.yaml

@@ -1,6 +1,8 @@
 esphome:
   on_boot:
     then:
+      - speaker.mute_on:
+      - speaker.mute_off:
       - if:
           condition: speaker.is_stopped
           then:

tests/components/speaker/test.esp32-idf.yaml

@@ -1,6 +1,8 @@
 esphome:
   on_boot:
     then:
+      - speaker.mute_on:
+      - speaker.mute_off:
       - if:
           condition: speaker.is_stopped
           then:
@@ -17,8 +19,17 @@ i2s_audio:
   i2s_bclk_pin: 17
   i2s_mclk_pin: 15
 
+i2c:
+  scl: 12
+  sda: 10
+
+audio_dac:
+  - platform: aic3204
+    id: internal_dac
+
 speaker:
   - platform: i2s_audio
-    id: speaker_id
+    id: speaker_with_audio_dac_id
+    audio_dac: internal_dac
     dac_type: external
-    i2s_dout_pin: 13
+    i2s_dout_pin: 14