diff --git a/esphome/components/audio/__init__.py b/esphome/components/audio/__init__.py
index 4ffdc401dc..31d3c39ffa 100644
--- a/esphome/components/audio/__init__.py
+++ b/esphome/components/audio/__init__.py
@@ -1,9 +1,121 @@
 import esphome.codegen as cg
 import esphome.config_validation as cv
+from esphome.const import CONF_BITS_PER_SAMPLE, CONF_NUM_CHANNELS, CONF_SAMPLE_RATE
+import esphome.final_validate as fv
 
 CODEOWNERS = ["@kahrendt"]
 audio_ns = cg.esphome_ns.namespace("audio")
 
+AudioFile = audio_ns.struct("AudioFile")
+AudioFileType = audio_ns.enum("AudioFileType", is_class=True)
+AUDIO_FILE_TYPE_ENUM = {
+    "NONE": AudioFileType.NONE,
+    "WAV": AudioFileType.WAV,
+    "MP3": AudioFileType.MP3,
+    "FLAC": AudioFileType.FLAC,
+}
+
+
+CONF_MIN_BITS_PER_SAMPLE = "min_bits_per_sample"
+CONF_MAX_BITS_PER_SAMPLE = "max_bits_per_sample"
+CONF_MIN_CHANNELS = "min_channels"
+CONF_MAX_CHANNELS = "max_channels"
+CONF_MIN_SAMPLE_RATE = "min_sample_rate"
+CONF_MAX_SAMPLE_RATE = "max_sample_rate"
+
+
 CONFIG_SCHEMA = cv.All(
     cv.Schema({}),
 )
+
+AUDIO_COMPONENT_SCHEMA = cv.Schema(
+    {
+        cv.Optional(CONF_BITS_PER_SAMPLE): cv.int_range(8, 32),
+        cv.Optional(CONF_NUM_CHANNELS): cv.int_range(1, 2),
+        cv.Optional(CONF_SAMPLE_RATE): cv.int_range(8000, 48000),
+    }
+)
+
+
+_UNDEF = object()
+
+
+def set_stream_limits(
+    min_bits_per_sample: int = _UNDEF,
+    max_bits_per_sample: int = _UNDEF,
+    min_channels: int = _UNDEF,
+    max_channels: int = _UNDEF,
+    min_sample_rate: int = _UNDEF,
+    max_sample_rate: int = _UNDEF,
+):
+    def set_limits_in_config(config):
+        if min_bits_per_sample is not _UNDEF:
+            config[CONF_MIN_BITS_PER_SAMPLE] = min_bits_per_sample
+        if max_bits_per_sample is not _UNDEF:
+            config[CONF_MAX_BITS_PER_SAMPLE] = max_bits_per_sample
+        if min_channels is not _UNDEF:
+            config[CONF_MIN_CHANNELS] = min_channels
+        if max_channels is not _UNDEF:
+            config[CONF_MAX_CHANNELS] = max_channels
+        if min_sample_rate is not _UNDEF:
+            config[CONF_MIN_SAMPLE_RATE] = min_sample_rate
+        if max_sample_rate is not _UNDEF:
+            config[CONF_MAX_SAMPLE_RATE] = max_sample_rate
+
+    return set_limits_in_config
+
+
+def final_validate_audio_schema(
+    name: str,
+    *,
+    audio_device: str,
+    bits_per_sample: int,
+    channels: int,
+    sample_rate: int,
+):
+    def validate_audio_compatiblity(audio_config):
+        audio_schema = {}
+
+        try:
+            cv.int_range(
+                min=audio_config.get(CONF_MIN_BITS_PER_SAMPLE),
+                max=audio_config.get(CONF_MAX_BITS_PER_SAMPLE),
+            )(bits_per_sample)
+        except cv.Invalid as exc:
+            raise cv.Invalid(
+                f"Invalid configuration for the {name} component. The {CONF_BITS_PER_SAMPLE} {str(exc)}"
+            ) from exc
+
+        try:
+            cv.int_range(
+                min=audio_config.get(CONF_MIN_CHANNELS),
+                max=audio_config.get(CONF_MAX_CHANNELS),
+            )(channels)
+        except cv.Invalid as exc:
+            raise cv.Invalid(
+                f"Invalid configuration for the {name} component. The {CONF_NUM_CHANNELS} {str(exc)}"
+            ) from exc
+
+        try:
+            cv.int_range(
+                min=audio_config.get(CONF_MIN_SAMPLE_RATE),
+                max=audio_config.get(CONF_MAX_SAMPLE_RATE),
+            )(sample_rate)
+            return cv.Schema(audio_schema, extra=cv.ALLOW_EXTRA)(audio_config)
+        except cv.Invalid as exc:
+            raise cv.Invalid(
+                f"Invalid configuration for the {name} component. The {CONF_SAMPLE_RATE} {str(exc)}"
+            ) from exc
+
+    return cv.Schema(
+        {
+            cv.Required(audio_device): fv.id_declaration_match_schema(
+                validate_audio_compatiblity
+            )
+        },
+        extra=cv.ALLOW_EXTRA,
+    )
+
+
+async def to_code(config):
+    cg.add_library("esphome/esp-audio-libs", "1.1.1")
diff --git a/esphome/components/audio/audio.cpp b/esphome/components/audio/audio.cpp
new file mode 100644
index 0000000000..2a58c38ac7
--- /dev/null
+++ b/esphome/components/audio/audio.cpp
@@ -0,0 +1,67 @@
+#include "audio.h"
+
+namespace esphome {
+namespace audio {
+
+// Euclidean's algorithm for finding the greatest common divisor
+static uint32_t gcd(uint32_t a, uint32_t b) {
+  while (b != 0) {
+    uint32_t t = b;
+    b = a % b;
+    a = t;
+  }
+  return a;
+}
+
+AudioStreamInfo::AudioStreamInfo(uint8_t bits_per_sample, uint8_t channels, uint32_t sample_rate)
+    : bits_per_sample_(bits_per_sample), channels_(channels), sample_rate_(sample_rate) {
+  this->ms_sample_rate_gcd_ = gcd(1000, this->sample_rate_);
+  this->bytes_per_sample_ = (this->bits_per_sample_ + 7) / 8;
+}
+
+uint32_t AudioStreamInfo::frames_to_microseconds(uint32_t frames) const {
+  return (frames * 1000000 + (this->sample_rate_ >> 1)) / this->sample_rate_;
+}
+
+uint32_t AudioStreamInfo::frames_to_milliseconds_with_remainder(uint32_t *total_frames) const {
+  uint32_t unprocessable_frames = *total_frames % (this->sample_rate_ / this->ms_sample_rate_gcd_);
+  uint32_t frames_for_ms_calculation = *total_frames - unprocessable_frames;
+
+  uint32_t playback_ms = (frames_for_ms_calculation * 1000) / this->sample_rate_;
+  *total_frames = unprocessable_frames;
+  return playback_ms;
+}
+
+bool AudioStreamInfo::operator==(const AudioStreamInfo &rhs) const {
+  return (this->bits_per_sample_ == rhs.get_bits_per_sample()) && (this->channels_ == rhs.get_channels()) &&
+         (this->sample_rate_ == rhs.get_sample_rate());
+}
+
+const char *audio_file_type_to_string(AudioFileType file_type) {
+  switch (file_type) {
+#ifdef USE_AUDIO_FLAC_SUPPORT
+    case AudioFileType::FLAC:
+      return "FLAC";
+#endif
+#ifdef USE_AUDIO_MP3_SUPPORT
+    case AudioFileType::MP3:
+      return "MP3";
+#endif
+    case AudioFileType::WAV:
+      return "WAV";
+    default:
+      return "unknown";
+  }
+}
+
+void scale_audio_samples(const int16_t *audio_samples, int16_t *output_buffer, int16_t scale_factor,
+                         size_t samples_to_scale) {
+  // Note the assembly dsps_mulc function has audio glitches if the input and output buffers are the same.
+  for (int i = 0; i < samples_to_scale; i++) {
+    int32_t acc = (int32_t) audio_samples[i] * (int32_t) scale_factor;
+    output_buffer[i] = (int16_t) (acc >> 15);
+  }
+}
+
+}  // namespace audio
+}  // namespace esphome
diff --git a/esphome/components/audio/audio.h b/esphome/components/audio/audio.h
index caf325cf54..6f0f1aaa46 100644
--- a/esphome/components/audio/audio.h
+++ b/esphome/components/audio/audio.h
@@ -1,21 +1,139 @@
 #pragma once
 
+#include "esphome/core/defines.h"
+
 #include <cstddef>
 #include <cstdint>
 
 namespace esphome {
 namespace audio {
 
-struct AudioStreamInfo {
-  bool operator==(const AudioStreamInfo &rhs) const {
-    return (channels == rhs.channels) && (bits_per_sample == rhs.bits_per_sample) && (sample_rate == rhs.sample_rate);
+class AudioStreamInfo {
+  /* Class to respresent important parameters of the audio stream that also provides helper function to convert between
+   * various audio related units.
+   *
+   *  - An audio sample represents a unit of audio for one channel.
+   *  - A frame represents a unit of audio with a sample for every channel.
+   *
+   * In gneneral, converting between bytes, samples, and frames shouldn't result in rounding errors so long as frames
+   * are used as the main unit when transferring audio data. Durations may result in rounding for certain sample rates;
+   * e.g., 44.1 KHz. The ``frames_to_milliseconds_with_remainder`` function should be used for accuracy, as it takes
+   * into account the remainder rather than just ignoring any rounding.
+   */
+ public:
+  AudioStreamInfo()
+      : AudioStreamInfo(16, 1, 16000){};  // Default values represent ESPHome's audio components historical values
+  AudioStreamInfo(uint8_t bits_per_sample, uint8_t channels, uint32_t sample_rate);
+
+  uint8_t get_bits_per_sample() const { return this->bits_per_sample_; }
+  uint8_t get_channels() const { return this->channels_; }
+  uint32_t get_sample_rate() const { return this->sample_rate_; }
+
+  /// @brief Convert bytes to duration in milliseconds.
+  /// @param bytes Number of bytes to convert
+  /// @return Duration in milliseconds that will store `bytes` bytes of audio. May round down for certain sample rates
+  ///         or values of `bytes`.
+  uint32_t bytes_to_ms(size_t bytes) const {
+    return bytes * 1000 / (this->sample_rate_ * this->bytes_per_sample_ * this->channels_);
   }
+
+  /// @brief Convert bytes to frames.
+  /// @param bytes Number of bytes to convert
+  /// @return Audio frames that will store `bytes` bytes.
+  uint32_t bytes_to_frames(size_t bytes) const { return (bytes / (this->bytes_per_sample_ * this->channels_)); }
+
+  /// @brief Convert bytes to samples.
+  /// @param bytes Number of bytes to convert
+  /// @return Audio samples that will store `bytes` bytes.
+  uint32_t bytes_to_samples(size_t bytes) const { return (bytes / this->bytes_per_sample_); }
+
+  /// @brief Converts frames to bytes.
+  /// @param frames Number of frames to convert.
+  /// @return Number of bytes that will store `frames` frames of audio.
+  size_t frames_to_bytes(uint32_t frames) const { return frames * this->bytes_per_sample_ * this->channels_; }
+
+  /// @brief Converts samples to bytes.
+  /// @param samples Number of samples to convert.
+  /// @return Number of bytes that will store `samples` samples of audio.
+  size_t samples_to_bytes(uint32_t samples) const { return samples * this->bytes_per_sample_; }
+
+  /// @brief Converts duration to frames.
+  /// @param ms Duration in milliseconds
+  /// @return Audio frames that will store `ms` milliseconds of audio.  May round down for certain sample rates.
+  uint32_t ms_to_frames(uint32_t ms) const { return (ms * this->sample_rate_) / 1000; }
+
+  /// @brief Converts duration to samples.
+  /// @param ms Duration in milliseconds
+  /// @return Audio samples that will store `ms` milliseconds of audio.  May round down for certain sample rates.
+  uint32_t ms_to_samples(uint32_t ms) const { return (ms * this->channels_ * this->sample_rate_) / 1000; }
+
+  /// @brief Converts duration to bytes. May round down for certain sample rates.
+  /// @param ms Duration in milliseconds
+  /// @return Bytes that will store `ms` milliseconds of audio.  May round down for certain sample rates.
+  size_t ms_to_bytes(uint32_t ms) const {
+    return (ms * this->bytes_per_sample_ * this->channels_ * this->sample_rate_) / 1000;
+  }
+
+  /// @brief Computes the duration, in microseconds, the given amount of frames represents.
+  /// @param frames Number of audio frames
+  /// @return Duration in microseconds `frames` respresents. May be slightly inaccurate due to integer divison rounding
+  ///         for certain sample rates.
+  uint32_t frames_to_microseconds(uint32_t frames) const;
+
+  /// @brief Computes the duration, in milliseconds, the given amount of frames represents. Avoids
+  /// accumulating rounding errors by updating `frames` with the remainder after converting.
+  /// @param frames Pointer to uint32_t with the number of audio frames. Replaced with the remainder.
+  /// @return Duration in milliseconds `frames` represents. Always less than or equal to the actual value due to
+  ///         rounding.
+  uint32_t frames_to_milliseconds_with_remainder(uint32_t *frames) const;
+
+  // Class comparison operators
+  bool operator==(const AudioStreamInfo &rhs) const;
   bool operator!=(const AudioStreamInfo &rhs) const { return !operator==(rhs); }
-  size_t get_bytes_per_sample() const { return bits_per_sample / 8; }
-  uint8_t channels = 1;
-  uint8_t bits_per_sample = 16;
-  uint32_t sample_rate = 16000;
+
+ protected:
+  uint8_t bits_per_sample_;
+  uint8_t channels_;
+  uint32_t sample_rate_;
+
+  // The greatest common divisor between 1000 ms = 1 second and the sample rate. Used to avoid accumulating error when
+  // converting from frames to duration. Computed at construction.
+  uint32_t ms_sample_rate_gcd_;
+
+  // Conversion factor derived from the number of bits per sample. Assumes audio data is aligned to the byte. Computed
+  // at construction.
+  size_t bytes_per_sample_;
 };
 
+enum class AudioFileType : uint8_t {
+  NONE = 0,
+#ifdef USE_AUDIO_FLAC_SUPPORT
+  FLAC,
+#endif
+#ifdef USE_AUDIO_MP3_SUPPORT
+  MP3,
+#endif
+  WAV,
+};
+
+struct AudioFile {
+  const uint8_t *data;
+  size_t length;
+  AudioFileType file_type;
+};
+
+/// @brief Helper function to convert file type to a const char string
+/// @param file_type
+/// @return const char pointer to the readable file type
+const char *audio_file_type_to_string(AudioFileType file_type);
+
+/// @brief Scales Q15 fixed point audio samples. Scales in place if audio_samples == output_buffer.
+/// @param audio_samples PCM int16 audio samples
+/// @param output_buffer Buffer to store the scaled samples
+/// @param scale_factor Q15 fixed point scaling factor
+/// @param samples_to_scale Number of samples to scale
+void scale_audio_samples(const int16_t *audio_samples, int16_t *output_buffer, int16_t scale_factor,
+                         size_t samples_to_scale);
+
 }  // namespace audio
 }  // namespace esphome
diff --git a/esphome/components/audio/audio_transfer_buffer.cpp b/esphome/components/audio/audio_transfer_buffer.cpp
new file mode 100644
index 0000000000..9b6067aac4
--- /dev/null
+++ b/esphome/components/audio/audio_transfer_buffer.cpp
@@ -0,0 +1,165 @@
+#include "audio_transfer_buffer.h"
+
+#ifdef USE_ESP32
+
+#include "esphome/core/helpers.h"
+
+namespace esphome {
+namespace audio {
+
+AudioTransferBuffer::~AudioTransferBuffer() { this->deallocate_buffer_(); };
+
+std::unique_ptr<AudioSinkTransferBuffer> AudioSinkTransferBuffer::create(size_t buffer_size) {
+  std::unique_ptr<AudioSinkTransferBuffer> sink_buffer = make_unique<AudioSinkTransferBuffer>();
+
+  if (!sink_buffer->allocate_buffer_(buffer_size)) {
+    return nullptr;
+  }
+
+  return sink_buffer;
+}
+
+std::unique_ptr<AudioSourceTransferBuffer> AudioSourceTransferBuffer::create(size_t buffer_size) {
+  std::unique_ptr<AudioSourceTransferBuffer> source_buffer = make_unique<AudioSourceTransferBuffer>();
+
+  if (!source_buffer->allocate_buffer_(buffer_size)) {
+    return nullptr;
+  }
+
+  return source_buffer;
+}
+
+size_t AudioTransferBuffer::free() const {
+  if (this->buffer_size_ == 0) {
+    return 0;
+  }
+  return this->buffer_size_ - (this->buffer_length_ - (this->data_start_ - this->buffer_));
+}
+
+void AudioTransferBuffer::decrease_buffer_length(size_t bytes) {
+  this->buffer_length_ -= bytes;
+  this->data_start_ += bytes;
+}
+
+void AudioTransferBuffer::increase_buffer_length(size_t bytes) { this->buffer_length_ += bytes; }
+
+void AudioTransferBuffer::clear_buffered_data() {
+  this->buffer_length_ = 0;
+  if (this->ring_buffer_.use_count() > 0) {
+    this->ring_buffer_->reset();
+  }
+}
+
+void AudioSinkTransferBuffer::clear_buffered_data() {
+  this->buffer_length_ = 0;
+  if (this->ring_buffer_.use_count() > 0) {
+    this->ring_buffer_->reset();
+  }
+#ifdef USE_SPEAKER
+  if (this->speaker_ != nullptr) {
+    this->speaker_->stop();
+  }
+#endif
+}
+
+bool AudioTransferBuffer::has_buffered_data() const {
+  if (this->ring_buffer_.use_count() > 0) {
+    return ((this->ring_buffer_->available() > 0) || (this->available() > 0));
+  }
+  return (this->available() > 0);
+}
+
+bool AudioTransferBuffer::reallocate(size_t new_buffer_size) {
+  if (this->buffer_length_ > 0) {
+    // Already has data in the buffer, fail
+    return false;
+  }
+  this->deallocate_buffer_();
+  return this->allocate_buffer_(new_buffer_size);
+}
+
+bool AudioTransferBuffer::allocate_buffer_(size_t buffer_size) {
+  this->buffer_size_ = buffer_size;
+
+  RAMAllocator<uint8_t> allocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
+
+  this->buffer_ = allocator.allocate(this->buffer_size_);
+  if (this->buffer_ == nullptr) {
+    return false;
+  }
+
+  this->data_start_ = this->buffer_;
+  this->buffer_length_ = 0;
+
+  return true;
+}
+
+void AudioTransferBuffer::deallocate_buffer_() {
+  if (this->buffer_ != nullptr) {
+    RAMAllocator<uint8_t> allocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
+    allocator.deallocate(this->buffer_, this->buffer_size_);
+    this->buffer_ = nullptr;
+    this->data_start_ = nullptr;
+  }
+
+  this->buffer_size_ = 0;
+  this->buffer_length_ = 0;
+}
+
+size_t AudioSourceTransferBuffer::transfer_data_from_source(TickType_t ticks_to_wait) {
+  // Shift data in buffer to start
+  if (this->buffer_length_ > 0) {
+    memmove(this->buffer_, this->data_start_, this->buffer_length_);
+  }
+  this->data_start_ = this->buffer_;
+
+  size_t bytes_to_read = this->free();
+  size_t bytes_read = 0;
+  if (bytes_to_read > 0) {
+    if (this->ring_buffer_.use_count() > 0) {
+      bytes_read = this->ring_buffer_->read((void *) this->get_buffer_end(), bytes_to_read, ticks_to_wait);
+    }
+
+    this->increase_buffer_length(bytes_read);
+  }
+  return bytes_read;
+}
+
+size_t AudioSinkTransferBuffer::transfer_data_to_sink(TickType_t ticks_to_wait) {
+  size_t bytes_written = 0;
+  if (this->available()) {
+#ifdef USE_SPEAKER
+    if (this->speaker_ != nullptr) {
+      bytes_written = this->speaker_->play(this->data_start_, this->available(), ticks_to_wait);
+    } else
+#endif
+        if (this->ring_buffer_.use_count() > 0) {
+      bytes_written =
+          this->ring_buffer_->write_without_replacement((void *) this->data_start_, this->available(), ticks_to_wait);
+    }
+
+    this->decrease_buffer_length(bytes_written);
+
+    // Shift unwritten data to the start of the buffer
+    memmove(this->buffer_, this->data_start_, this->buffer_length_);
+    this->data_start_ = this->buffer_;
+  }
+  return bytes_written;
+}
+
+bool AudioSinkTransferBuffer::has_buffered_data() const {
+#ifdef USE_SPEAKER
+  if (this->speaker_ != nullptr) {
+    return (this->speaker_->has_buffered_data() || (this->available() > 0));
+  }
+#endif
+  if (this->ring_buffer_.use_count() > 0) {
+    return ((this->ring_buffer_->available() > 0) || (this->available() > 0));
+  }
+  return (this->available() > 0);
+}
+
+}  // namespace audio
+}  // namespace esphome
+
+#endif
diff --git a/esphome/components/audio/audio_transfer_buffer.h b/esphome/components/audio/audio_transfer_buffer.h
new file mode 100644
index 0000000000..4e461db56d
--- /dev/null
+++ b/esphome/components/audio/audio_transfer_buffer.h
@@ -0,0 +1,139 @@
+#pragma once
+
+#ifdef USE_ESP32
+#include "esphome/core/defines.h"
+#include "esphome/core/ring_buffer.h"
+
+#ifdef USE_SPEAKER
+#include "esphome/components/speaker/speaker.h"
+#endif
+
+#include "esp_err.h"
+
+#include <freertos/FreeRTOS.h>
+
+namespace esphome {
+namespace audio {
+
+class AudioTransferBuffer {
+  /*
+   * @brief Class that facilitates tranferring data between a buffer and an audio source or sink.
+   * The transfer buffer is a typical C array that temporarily holds data for processing in other audio components.
+   * Both sink and source transfer buffers can use a ring buffer as the sink/source.
+   *   - The ring buffer is stored in a shared_ptr, so destroying the transfer buffer object will release ownership.
+   */
+ public:
+  /// @brief Destructor that deallocates the transfer buffer
+  ~AudioTransferBuffer();
+
+  /// @brief Returns a pointer to the start of the transfer buffer where available() bytes of exisiting data can be read
+  uint8_t *get_buffer_start() const { return this->data_start_; }
+
+  /// @brief Returns a pointer to the end of the transfer buffer where free() bytes of new data can be written
+  uint8_t *get_buffer_end() const { return this->data_start_ + this->buffer_length_; }
+
+  /// @brief Updates the internal state of the transfer buffer. This should be called after reading data
+  /// @param bytes The number of bytes consumed/read
+  void decrease_buffer_length(size_t bytes);
+
+  /// @brief Updates the internal state of the transfer buffer. This should be called after writing data
+  /// @param bytes The number of bytes written
+  void increase_buffer_length(size_t bytes);
+
+  /// @brief Returns the transfer buffer's currently available bytes to read
+  size_t available() const { return this->buffer_length_; }
+
+  /// @brief Returns the transfer buffers allocated bytes
+  size_t capacity() const { return this->buffer_size_; }
+
+  /// @brief Returns the transfer buffer's currrently free bytes available to write
+  size_t free() const;
+
+  /// @brief Clears data in the transfer buffer and, if possible, the source/sink.
+  virtual void clear_buffered_data();
+
+  /// @brief Tests if there is any data in the tranfer buffer or the source/sink.
+  /// @return True if there is data, false otherwise.
+  virtual bool has_buffered_data() const;
+
+  bool reallocate(size_t new_buffer_size);
+
+ protected:
+  /// @brief Allocates the transfer buffer in external memory, if available.
+  /// @return True is successful, false otherwise.
+  bool allocate_buffer_(size_t buffer_size);
+
+  /// @brief Deallocates the buffer and resets the class variables.
+  void deallocate_buffer_();
+
+  // A possible source or sink for the transfer buffer
+  std::shared_ptr<RingBuffer> ring_buffer_;
+
+  uint8_t *buffer_{nullptr};
+  uint8_t *data_start_{nullptr};
+
+  size_t buffer_size_{0};
+  size_t buffer_length_{0};
+};
+
+class AudioSinkTransferBuffer : public AudioTransferBuffer {
+  /*
+   * @brief A class that implements a transfer buffer for audio sinks.
+   * Supports writing processed data in the transfer buffer to a ring buffer or a speaker component.
+   */
+ public:
+  /// @brief Creates a new sink transfer buffer.
+  /// @param buffer_size Size of the transfer buffer in bytes.
+  /// @return unique_ptr if successfully allocated, nullptr otherwise
+  static std::unique_ptr<AudioSinkTransferBuffer> create(size_t buffer_size);
+
+  /// @brief Writes any available data in the transfer buffer to the sink.
+  /// @param ticks_to_wait FreeRTOS ticks to block while waiting for the sink to have enough space
+  /// @return Number of bytes written
+  size_t transfer_data_to_sink(TickType_t ticks_to_wait);
+
+  /// @brief Adds a ring buffer as the transfer buffer's sink.
+  /// @param ring_buffer weak_ptr to the allocated ring buffer
+  void set_sink(const std::weak_ptr<RingBuffer> &ring_buffer) { this->ring_buffer_ = ring_buffer.lock(); }
+
+#ifdef USE_SPEAKER
+  /// @brief Adds a speaker as the transfer buffer's sink.
+  /// @param speaker Pointer to the speaker component
+  void set_sink(speaker::Speaker *speaker) { this->speaker_ = speaker; }
+#endif
+
+  void clear_buffered_data() override;
+
+  bool has_buffered_data() const override;
+
+ protected:
+#ifdef USE_SPEAKER
+  speaker::Speaker *speaker_{nullptr};
+#endif
+};
+
+class AudioSourceTransferBuffer : public AudioTransferBuffer {
+  /*
+   * @brief A class that implements a transfer buffer for audio sources.
+   * Supports reading audio data from a ring buffer into the transfer buffer for processing.
+   */
+ public:
+  /// @brief Creates a new source transfer buffer.
+  /// @param buffer_size Size of the transfer buffer in bytes.
+  /// @return unique_ptr if successfully allocated, nullptr otherwise
+  static std::unique_ptr<AudioSourceTransferBuffer> create(size_t buffer_size);
+
+  /// @brief Reads any available data from the sink into the transfer buffer.
+  /// @param ticks_to_wait FreeRTOS ticks to block while waiting for the source to have enough data
+  /// @return Number of bytes read
+  size_t transfer_data_from_source(TickType_t ticks_to_wait);
+
+  /// @brief Adds a ring buffer as the transfer buffer's source.
+  /// @param ring_buffer weak_ptr to the allocated ring buffer
+  void set_source(const std::weak_ptr<RingBuffer> &ring_buffer) { this->ring_buffer_ = ring_buffer.lock(); };
+};
+
+}  // namespace audio
+}  // namespace esphome
+
+#endif
diff --git a/esphome/components/esp32/gpio.h b/esphome/components/esp32/gpio.h
index 23b723e0b4..d69ac1c493 100644
--- a/esphome/components/esp32/gpio.h
+++ b/esphome/components/esp32/gpio.h
@@ -13,6 +13,7 @@ class ESP32InternalGPIOPin : public InternalGPIOPin {
   void set_inverted(bool inverted) { inverted_ = inverted; }
   void set_drive_strength(gpio_drive_cap_t drive_strength) { drive_strength_ = drive_strength; }
   void set_flags(gpio::Flags flags) { flags_ = flags; }
+
   void setup() override;
   void pin_mode(gpio::Flags flags) override;
   bool digital_read() override;
@@ -21,6 +22,7 @@ class ESP32InternalGPIOPin : public InternalGPIOPin {
   void detach_interrupt() const override;
   ISRInternalGPIOPin to_isr() const override;
   uint8_t get_pin() const override { return (uint8_t) pin_; }
+  gpio::Flags get_flags() const override { return flags_; }
   bool is_inverted() const override { return inverted_; }
 
  protected:
diff --git a/esphome/components/esp8266/gpio.h b/esphome/components/esp8266/gpio.h
index 0474d0baa6..dd6407885e 100644
--- a/esphome/components/esp8266/gpio.h
+++ b/esphome/components/esp8266/gpio.h
@@ -22,6 +22,7 @@ class ESP8266GPIOPin : public InternalGPIOPin {
   void detach_interrupt() const override;
   ISRInternalGPIOPin to_isr() const override;
   uint8_t get_pin() const override { return pin_; }
+  gpio::Flags get_flags() const override { return flags_; }
   bool is_inverted() const override { return inverted_; }
 
  protected:
diff --git a/esphome/components/host/gpio.h b/esphome/components/host/gpio.h
index c0920467d6..a60d535912 100644
--- a/esphome/components/host/gpio.h
+++ b/esphome/components/host/gpio.h
@@ -21,6 +21,7 @@ class HostGPIOPin : public InternalGPIOPin {
   void detach_interrupt() const override;
   ISRInternalGPIOPin to_isr() const override;
   uint8_t get_pin() const override { return pin_; }
+  gpio::Flags get_flags() const override { return flags_; }
   bool is_inverted() const override { return inverted_; }
 
  protected:
diff --git a/esphome/components/i2s_audio/speaker/__init__.py b/esphome/components/i2s_audio/speaker/__init__.py
index eb6bb329ad..aa3b50d336 100644
--- a/esphome/components/i2s_audio/speaker/__init__.py
+++ b/esphome/components/i2s_audio/speaker/__init__.py
@@ -1,20 +1,25 @@
 from esphome import pins
 import esphome.codegen as cg
-from esphome.components import esp32, speaker
+from esphome.components import audio, esp32, speaker
 import esphome.config_validation as cv
 from esphome.const import (
+    CONF_BITS_PER_SAMPLE,
     CONF_BUFFER_DURATION,
     CONF_CHANNEL,
     CONF_ID,
     CONF_MODE,
     CONF_NEVER,
+    CONF_NUM_CHANNELS,
+    CONF_SAMPLE_RATE,
     CONF_TIMEOUT,
 )
 
 from .. import (
     CONF_I2S_DOUT_PIN,
+    CONF_I2S_MODE,
     CONF_LEFT,
     CONF_MONO,
+    CONF_PRIMARY,
     CONF_RIGHT,
     CONF_STEREO,
     I2SAudioOut,
@@ -58,7 +63,41 @@ I2C_COMM_FMT_OPTIONS = {
 NO_INTERNAL_DAC_VARIANTS = [esp32.const.VARIANT_ESP32S2]
 
 
-def validate_esp32_variant(config):
+def _set_num_channels_from_config(config):
+    if config[CONF_CHANNEL] in (CONF_MONO, CONF_LEFT, CONF_RIGHT):
+        config[CONF_NUM_CHANNELS] = 1
+    else:
+        config[CONF_NUM_CHANNELS] = 2
+
+    return config
+
+
+def _set_stream_limits(config):
+    if config[CONF_I2S_MODE] == CONF_PRIMARY:
+        # Primary mode has modifiable stream settings
+        audio.set_stream_limits(
+            min_bits_per_sample=8,
+            max_bits_per_sample=32,
+            min_channels=1,
+            max_channels=2,
+            min_sample_rate=16000,
+            max_sample_rate=48000,
+        )(config)
+    else:
+        # Secondary mode has unmodifiable max bits per sample and min/max sample rates
+        audio.set_stream_limits(
+            min_bits_per_sample=8,
+            max_bits_per_sample=config.get(CONF_BITS_PER_SAMPLE),
+            min_channels=1,
+            max_channels=2,
+            min_sample_rate=config.get(CONF_SAMPLE_RATE),
+            max_sample_rate=config.get(CONF_SAMPLE_RATE),
+        )
+
+    return config
+
+
+def _validate_esp32_variant(config):
     if config[CONF_DAC_TYPE] != "internal":
         return config
     variant = esp32.get_esp32_variant()
@@ -90,6 +129,7 @@ BASE_SCHEMA = (
     .extend(cv.COMPONENT_SCHEMA)
 )
 
+
 CONFIG_SCHEMA = cv.All(
     cv.typed_schema(
         {
@@ -111,7 +151,9 @@ CONFIG_SCHEMA = cv.All(
         },
         key=CONF_DAC_TYPE,
     ),
-    validate_esp32_variant,
+    _validate_esp32_variant,
+    _set_num_channels_from_config,
+    _set_stream_limits,
 )
 
 
diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp
index 46f1b00d05..2a3fa9f5f3 100644
--- a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp
+++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp
@@ -148,9 +148,11 @@ void I2SAudioSpeaker::loop() {
     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);
+             this->audio_stream_info_.get_sample_rate(), this->audio_stream_info_.get_channels(),
+             this->audio_stream_info_.get_bits_per_sample());
   }
+
+  xEventGroupClearBits(this->event_group_, ALL_ERR_ESP_BITS);
 }
 
 void I2SAudioSpeaker::set_volume(float volume) {
@@ -201,6 +203,12 @@ size_t I2SAudioSpeaker::play(const uint8_t *data, size_t length, TickType_t tick
     this->start();
   }
 
+  if ((this->state_ != speaker::STATE_RUNNING) || (this->audio_ring_buffer_.use_count() == 1)) {
+    // Unable to write data to a running speaker, so delay the max amount of time so it can get ready
+    vTaskDelay(ticks_to_wait);
+    ticks_to_wait = 0;
+  }
+
   size_t bytes_written = 0;
   if ((this->state_ == speaker::STATE_RUNNING) && (this->audio_ring_buffer_.use_count() == 1)) {
     // Only one owner of the ring buffer (the speaker task), so the ring buffer is allocated and no other components are
@@ -223,6 +231,8 @@ bool I2SAudioSpeaker::has_buffered_data() const {
 
 void I2SAudioSpeaker::speaker_task(void *params) {
   I2SAudioSpeaker *this_speaker = (I2SAudioSpeaker *) params;
+  this_speaker->task_created_ = true;
+
   uint32_t event_group_bits =
       xEventGroupWaitBits(this_speaker->event_group_,
                           SpeakerEventGroupBits::COMMAND_START | SpeakerEventGroupBits::COMMAND_STOP |
@@ -240,19 +250,20 @@ void I2SAudioSpeaker::speaker_task(void *params) {
 
   audio::AudioStreamInfo audio_stream_info = this_speaker->audio_stream_info_;
 
-  const uint32_t bytes_per_ms =
-      audio_stream_info.channels * audio_stream_info.get_bytes_per_sample() * audio_stream_info.sample_rate / 1000;
+  const uint32_t dma_buffers_duration_ms = DMA_BUFFER_DURATION_MS * DMA_BUFFERS_COUNT;
+  // Ensure ring buffer duration is at least the duration of all DMA buffers
+  const uint32_t ring_buffer_duration = std::max(dma_buffers_duration_ms, this_speaker->buffer_duration_ms_);
 
-  const size_t dma_buffers_size = DMA_BUFFERS_COUNT * DMA_BUFFER_DURATION_MS * bytes_per_ms;
+  // The DMA buffers may have more bits per sample, so calculate buffer sizes based in the input audio stream info
+  const size_t data_buffer_size = audio_stream_info.ms_to_bytes(dma_buffers_duration_ms);
+  const size_t ring_buffer_size = audio_stream_info.ms_to_bytes(ring_buffer_duration);
 
-  // Ensure ring buffer is at least as large as the total size of the DMA buffers
-  const size_t ring_buffer_size =
-      std::max((uint32_t) dma_buffers_size, this_speaker->buffer_duration_ms_ * bytes_per_ms);
+  const size_t single_dma_buffer_input_size = data_buffer_size / DMA_BUFFERS_COUNT;
 
-  if (this_speaker->send_esp_err_to_event_group_(this_speaker->allocate_buffers_(dma_buffers_size, ring_buffer_size))) {
+  if (this_speaker->send_esp_err_to_event_group_(this_speaker->allocate_buffers_(data_buffer_size, ring_buffer_size))) {
     // Failed to allocate buffers
     xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_ESP_NO_MEM);
-    this_speaker->delete_task_(dma_buffers_size);
+    this_speaker->delete_task_(data_buffer_size);
   }
 
   if (!this_speaker->send_esp_err_to_event_group_(this_speaker->start_i2s_driver_(audio_stream_info))) {
@@ -262,20 +273,25 @@ void I2SAudioSpeaker::speaker_task(void *params) {
     uint32_t last_data_received_time = millis();
     bool tx_dma_underflow = false;
 
-    while (!this_speaker->timeout_.has_value() ||
+    this_speaker->accumulated_frames_written_ = 0;
+
+    // Keep looping if paused, there is no timeout configured, or data was received more recently than the configured
+    // timeout
+    while (this_speaker->pause_state_ || !this_speaker->timeout_.has_value() ||
            (millis() - last_data_received_time) <= this_speaker->timeout_.value()) {
       event_group_bits = xEventGroupGetBits(this_speaker->event_group_);
 
       if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP) {
+        xEventGroupClearBits(this_speaker->event_group_, SpeakerEventGroupBits::COMMAND_STOP);
         break;
       }
       if (event_group_bits & SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY) {
+        xEventGroupClearBits(this_speaker->event_group_, SpeakerEventGroupBits::COMMAND_STOP_GRACEFULLY);
         stop_gracefully = true;
       }
 
       if (this_speaker->audio_stream_info_ != audio_stream_info) {
-        // Audio stream info has changed, stop the speaker task so it will restart with the proper settings.
-
+        // Audio stream info changed, stop the speaker task so it will restart with the proper settings.
         break;
       }
 
@@ -286,33 +302,64 @@ void I2SAudioSpeaker::speaker_task(void *params) {
         }
       }
 
-      size_t bytes_to_read = dma_buffers_size;
-      size_t bytes_read = this_speaker->audio_ring_buffer_->read((void *) this_speaker->data_buffer_, bytes_to_read,
+      if (this_speaker->pause_state_) {
+        // Pause state is accessed atomically, so thread safe
+        // Delay so the task can yields, then skip transferring audio data
+        delay(TASK_DELAY_MS);
+        continue;
+      }
+
+      size_t bytes_read = this_speaker->audio_ring_buffer_->read((void *) this_speaker->data_buffer_, data_buffer_size,
                                                                  pdMS_TO_TICKS(TASK_DELAY_MS));
 
       if (bytes_read > 0) {
-        size_t bytes_written = 0;
-
-        if ((audio_stream_info.bits_per_sample == 16) && (this_speaker->q15_volume_factor_ < INT16_MAX)) {
+        if ((audio_stream_info.get_bits_per_sample() == 16) && (this_speaker->q15_volume_factor_ < INT16_MAX)) {
           // Scale samples by the volume factor in place
           q15_multiplication((int16_t *) this_speaker->data_buffer_, (int16_t *) this_speaker->data_buffer_,
                              bytes_read / sizeof(int16_t), this_speaker->q15_volume_factor_);
         }
 
-        if (audio_stream_info.bits_per_sample == (uint8_t) this_speaker->bits_per_sample_) {
-          i2s_write(this_speaker->parent_->get_port(), this_speaker->data_buffer_, bytes_read, &bytes_written,
-                    portMAX_DELAY);
-        } else if (audio_stream_info.bits_per_sample < (uint8_t) this_speaker->bits_per_sample_) {
-          i2s_write_expand(this_speaker->parent_->get_port(), this_speaker->data_buffer_, bytes_read,
-                           audio_stream_info.bits_per_sample, this_speaker->bits_per_sample_, &bytes_written,
-                           portMAX_DELAY);
-        }
+        // Write the audio data to a single DMA buffer at a time to reduce latency for the audio duration played
+        // callback.
+        const uint32_t batches = (bytes_read + single_dma_buffer_input_size - 1) / single_dma_buffer_input_size;
 
-        if (bytes_written != bytes_read) {
-          xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_ESP_INVALID_SIZE);
+        for (uint32_t i = 0; i < batches; ++i) {
+          size_t bytes_written = 0;
+          size_t bytes_to_write = std::min(single_dma_buffer_input_size, bytes_read);
+
+          if (audio_stream_info.get_bits_per_sample() == (uint8_t) this_speaker->bits_per_sample_) {
+            i2s_write(this_speaker->parent_->get_port(), this_speaker->data_buffer_ + i * single_dma_buffer_input_size,
+                      bytes_to_write, &bytes_written, pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS * 5));
+          } else if (audio_stream_info.get_bits_per_sample() < (uint8_t) this_speaker->bits_per_sample_) {
+            i2s_write_expand(this_speaker->parent_->get_port(),
+                             this_speaker->data_buffer_ + i * single_dma_buffer_input_size, bytes_to_write,
+                             audio_stream_info.get_bits_per_sample(), this_speaker->bits_per_sample_, &bytes_written,
+                             pdMS_TO_TICKS(DMA_BUFFER_DURATION_MS * 5));
+          }
+
+          uint32_t write_timestamp = micros();
+
+          if (bytes_written != bytes_to_write) {
+            xEventGroupSetBits(this_speaker->event_group_, SpeakerEventGroupBits::ERR_ESP_INVALID_SIZE);
+          }
+
+          bytes_read -= bytes_written;
+
+          this_speaker->accumulated_frames_written_ += audio_stream_info.bytes_to_frames(bytes_written);
+          const uint32_t new_playback_ms =
+              audio_stream_info.frames_to_milliseconds_with_remainder(&this_speaker->accumulated_frames_written_);
+          const uint32_t remainder_us =
+              audio_stream_info.frames_to_microseconds(this_speaker->accumulated_frames_written_);
+
+          uint32_t pending_frames =
+              audio_stream_info.bytes_to_frames(bytes_read + this_speaker->audio_ring_buffer_->available());
+          const uint32_t pending_ms = audio_stream_info.frames_to_milliseconds_with_remainder(&pending_frames);
+
+          this_speaker->audio_output_callback_(new_playback_ms, remainder_us, pending_ms, write_timestamp);
+
+          tx_dma_underflow = false;
+          last_data_received_time = millis();
         }
-        tx_dma_underflow = false;
-        last_data_received_time = millis();
       } else {
         // No data received
         if (stop_gracefully && tx_dma_underflow) {
@@ -328,7 +375,7 @@ void I2SAudioSpeaker::speaker_task(void *params) {
     this_speaker->parent_->unlock();
   }
 
-  this_speaker->delete_task_(dma_buffers_size);
+  this_speaker->delete_task_(data_buffer_size);
 }
 
 void I2SAudioSpeaker::start() {
@@ -337,16 +384,15 @@ void I2SAudioSpeaker::start() {
   if ((this->state_ == speaker::STATE_STARTING) || (this->state_ == speaker::STATE_RUNNING))
     return;
 
-  if (this->speaker_task_handle_ == nullptr) {
+  if (!this->task_created_ && (this->speaker_task_handle_ == nullptr)) {
     xTaskCreate(I2SAudioSpeaker::speaker_task, "speaker_task", TASK_STACK_SIZE, (void *) this, TASK_PRIORITY,
                 &this->speaker_task_handle_);
-  }
 
-  if (this->speaker_task_handle_ != nullptr) {
-    xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_START);
-    this->task_created_ = true;
-  } else {
-    xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START);
+    if (this->speaker_task_handle_ != nullptr) {
+      xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::COMMAND_START);
+    } else {
+      xEventGroupSetBits(this->event_group_, SpeakerEventGroupBits::ERR_TASK_FAILED_TO_START);
+    }
   }
 }
 
@@ -416,12 +462,12 @@ esp_err_t I2SAudioSpeaker::allocate_buffers_(size_t data_buffer_size, size_t rin
 }
 
 esp_err_t I2SAudioSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_stream_info) {
-  if ((this->i2s_mode_ & I2S_MODE_SLAVE) && (this->sample_rate_ != audio_stream_info.sample_rate)) {  // NOLINT
-    //  Can't reconfigure I2S bus, so the sample rate must match the configured value
+  if ((this->i2s_mode_ & I2S_MODE_SLAVE) && (this->sample_rate_ != audio_stream_info.get_sample_rate())) {  // NOLINT
+    // Can't reconfigure I2S bus, so the sample rate must match the configured value
     return ESP_ERR_NOT_SUPPORTED;
   }
 
-  if ((i2s_bits_per_sample_t) audio_stream_info.bits_per_sample > this->bits_per_sample_) {
+  if ((i2s_bits_per_sample_t) audio_stream_info.get_bits_per_sample() > this->bits_per_sample_) {
     // Currently can't handle the case when the incoming audio has more bits per sample than the configured value
     return ESP_ERR_NOT_SUPPORTED;
   }
@@ -432,21 +478,21 @@ esp_err_t I2SAudioSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_strea
 
   i2s_channel_fmt_t channel = this->channel_;
 
-  if (audio_stream_info.channels == 1) {
+  if (audio_stream_info.get_channels() == 1) {
     if (this->channel_ == I2S_CHANNEL_FMT_ONLY_LEFT) {
       channel = I2S_CHANNEL_FMT_ONLY_LEFT;
     } else {
       channel = I2S_CHANNEL_FMT_ONLY_RIGHT;
     }
-  } else if (audio_stream_info.channels == 2) {
+  } else if (audio_stream_info.get_channels() == 2) {
     channel = I2S_CHANNEL_FMT_RIGHT_LEFT;
   }
 
-  int dma_buffer_length = DMA_BUFFER_DURATION_MS * this->sample_rate_ / 1000;
+  int dma_buffer_length = audio_stream_info.ms_to_frames(DMA_BUFFER_DURATION_MS);
 
   i2s_driver_config_t config = {
     .mode = (i2s_mode_t) (this->i2s_mode_ | I2S_MODE_TX),
-    .sample_rate = audio_stream_info.sample_rate,
+    .sample_rate = audio_stream_info.get_sample_rate(),
     .bits_per_sample = this->bits_per_sample_,
     .channel_format = channel,
     .communication_format = this->i2s_comm_fmt_,
@@ -504,7 +550,7 @@ esp_err_t I2SAudioSpeaker::start_i2s_driver_(audio::AudioStreamInfo &audio_strea
 }
 
 void I2SAudioSpeaker::delete_task_(size_t buffer_size) {
-  this->audio_ring_buffer_.reset();  // Releases onwership of the shared_ptr
+  this->audio_ring_buffer_.reset();  // Releases ownership of the shared_ptr
 
   if (this->data_buffer_ != nullptr) {
     ExternalRAMAllocator<uint8_t> allocator(ExternalRAMAllocator<uint8_t>::ALLOW_FAILURE);
diff --git a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h
index d706deb0f4..7b14a57aac 100644
--- a/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h
+++ b/esphome/components/i2s_audio/speaker/i2s_audio_speaker.h
@@ -40,6 +40,9 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp
   void stop() override;
   void finish() override;
 
+  void set_pause_state(bool pause_state) override { this->pause_state_ = pause_state; }
+  bool get_pause_state() const override { return this->pause_state_; }
+
   /// @brief Plays the provided audio data.
   /// Starts the speaker task, if necessary. Writes the audio data to the ring buffer.
   /// @param data Audio data in the format set by the parent speaker classes ``set_audio_stream_info`` method.
@@ -121,13 +124,18 @@ class I2SAudioSpeaker : public I2SAudioOut, public speaker::Speaker, public Comp
   uint8_t dout_pin_;
 
   bool task_created_{false};
+  bool pause_state_{false};
 
   int16_t q15_volume_factor_{INT16_MAX};
 
+  size_t bytes_written_{0};
+
 #if SOC_I2S_SUPPORTS_DAC
   i2s_dac_mode_t internal_dac_mode_{I2S_DAC_CHANNEL_DISABLE};
 #endif
   i2s_comm_format_t i2s_comm_fmt_;
+
+  uint32_t accumulated_frames_written_{0};
 };
 
 }  // namespace i2s_audio
diff --git a/esphome/components/libretiny/gpio_arduino.h b/esphome/components/libretiny/gpio_arduino.h
index a43ed28c5e..9adc425a41 100644
--- a/esphome/components/libretiny/gpio_arduino.h
+++ b/esphome/components/libretiny/gpio_arduino.h
@@ -20,6 +20,7 @@ class ArduinoInternalGPIOPin : public InternalGPIOPin {
   void detach_interrupt() const override;
   ISRInternalGPIOPin to_isr() const override;
   uint8_t get_pin() const override { return pin_; }
+  gpio::Flags get_flags() const override { return flags_; }
   bool is_inverted() const override { return inverted_; }
 
  protected:
diff --git a/esphome/components/rp2040/gpio.h b/esphome/components/rp2040/gpio.h
index ef9500d5dd..9bc66d9e4b 100644
--- a/esphome/components/rp2040/gpio.h
+++ b/esphome/components/rp2040/gpio.h
@@ -22,6 +22,7 @@ class RP2040GPIOPin : public InternalGPIOPin {
   void detach_interrupt() const override;
   ISRInternalGPIOPin to_isr() const override;
   uint8_t get_pin() const override { return pin_; }
+  gpio::Flags get_flags() const override { return flags_; }
   bool is_inverted() const override { return inverted_; }
 
  protected:
diff --git a/esphome/components/speaker/__init__.py b/esphome/components/speaker/__init__.py
index 948fe4b534..2ac1ca0cb9 100644
--- a/esphome/components/speaker/__init__.py
+++ b/esphome/components/speaker/__init__.py
@@ -1,7 +1,6 @@
 from esphome import automation
-from esphome.automation import maybe_simple_id
 import esphome.codegen as cg
-from esphome.components import audio_dac
+from esphome.components import audio, audio_dac
 import esphome.config_validation as cv
 from esphome.const import CONF_DATA, CONF_ID, CONF_VOLUME
 from esphome.core import CORE
@@ -54,13 +53,15 @@ async def register_speaker(var, config):
     await setup_speaker_core_(var, config)
 
 
-SPEAKER_SCHEMA = cv.Schema(
+SPEAKER_SCHEMA = cv.Schema.extend(audio.AUDIO_COMPONENT_SCHEMA).extend(
     {
         cv.Optional(CONF_AUDIO_DAC): cv.use_id(audio_dac.AudioDac),
     }
 )
 
-SPEAKER_AUTOMATION_SCHEMA = maybe_simple_id({cv.GenerateID(): cv.use_id(Speaker)})
+SPEAKER_AUTOMATION_SCHEMA = automation.maybe_simple_id(
+    {cv.GenerateID(): cv.use_id(Speaker)}
+)
 
 
 async def speaker_action(config, action_id, template_arg, args):
diff --git a/esphome/components/speaker/speaker.h b/esphome/components/speaker/speaker.h
index 96843e2d5a..74c4822eca 100644
--- a/esphome/components/speaker/speaker.h
+++ b/esphome/components/speaker/speaker.h
@@ -9,6 +9,7 @@
 #endif
 
 #include "esphome/core/defines.h"
+#include "esphome/core/helpers.h"
 
 #include "esphome/components/audio/audio.h"
 #ifdef USE_AUDIO_DAC
@@ -56,6 +57,10 @@ class Speaker {
   // When finish() is not implemented on the platform component it should just do a normal stop.
   virtual void finish() { this->stop(); }
 
+  // Pauses processing incoming audio. Needs to be implemented specifically per speaker component
+  virtual void set_pause_state(bool pause_state) {}
+  virtual bool get_pause_state() const { return false; }
+
   virtual bool has_buffered_data() const = 0;
 
   bool is_running() const { return this->state_ == STATE_RUNNING; }
@@ -95,6 +100,19 @@ class Speaker {
     this->audio_stream_info_ = audio_stream_info;
   }
 
+  audio::AudioStreamInfo &get_audio_stream_info() { return this->audio_stream_info_; }
+
+  /// Callback function for sending the duration of the audio written to the speaker since the last callback.
+  /// Parameters:
+  ///   - Duration in milliseconds. Never rounded and should always be less than or equal to the actual duration.
+  ///   - Remainder duration in microseconds. Rounded duration after subtracting the previous parameter from the actual
+  ///     duration.
+  ///   - Duration of remaining, unwritten audio buffered in the speaker in milliseconds.
+  ///   - System time in microseconds when the last write was completed.
+  void add_audio_output_callback(std::function<void(uint32_t, uint32_t, uint32_t, uint32_t)> &&callback) {
+    this->audio_output_callback_.add(std::move(callback));
+  }
+
  protected:
   State state_{STATE_STOPPED};
   audio::AudioStreamInfo audio_stream_info_;
@@ -104,6 +122,8 @@ class Speaker {
 #ifdef USE_AUDIO_DAC
   audio_dac::AudioDac *audio_dac_{nullptr};
 #endif
+
+  CallbackManager<void(uint32_t, uint32_t, uint32_t, uint32_t)> audio_output_callback_{};
 };
 
 }  // namespace speaker
diff --git a/esphome/core/gpio.h b/esphome/core/gpio.h
index 1b6f2ba1e6..19d57a0af8 100644
--- a/esphome/core/gpio.h
+++ b/esphome/core/gpio.h
@@ -53,6 +53,16 @@ class GPIOPin {
 
   virtual void pin_mode(gpio::Flags flags) = 0;
 
+  /**
+   * @brief Retrieve GPIO pin flags.
+   *
+   * @note This is currently optional to limit changes but will be mandatory in a future update.
+   * It is primarily applied to internal pins for now.
+   *
+   * @return The GPIO flags describing the pin mode and properties. Returns `gpio::Flags::FLAG_NONE` if not overridden.
+   */
+  virtual gpio::Flags get_flags() const { return gpio::Flags::FLAG_NONE; }
+
   virtual bool digital_read() = 0;
 
   virtual void digital_write(bool value) = 0;
diff --git a/platformio.ini b/platformio.ini
index e91c06d86e..cf11139b73 100644
--- a/platformio.ini
+++ b/platformio.ini
@@ -127,7 +127,8 @@ lib_deps =
     ESPmDNS                              ; mdns (Arduino built-in)
     DNSServer                            ; captive_portal (Arduino built-in)
     esphome/ESP32-audioI2S@2.0.7         ; i2s_audio
-    droscy/esp_wireguard@0.4.2          ; wireguard
+    droscy/esp_wireguard@0.4.2           ; wireguard
+    esphome/esp-audio-libs@1.1.1         ; audio
 
 build_flags =
     ${common:arduino.build_flags}
@@ -148,6 +149,7 @@ lib_deps =
     ${common:idf.lib_deps}
     droscy/esp_wireguard@0.4.2              ; wireguard
     kahrendt/ESPMicroSpeechFeatures@1.1.0   ; micro_wake_word
+    esphome/esp-audio-libs@1.1.1            ; audio
 build_flags =
     ${common:idf.build_flags}
     -Wno-nonnull-compare