diff --git a/esphome/components/esp32_ble/ble.h b/esphome/components/esp32_ble/ble.h
index 0674d293e1..6381e12448 100644
--- a/esphome/components/esp32_ble/ble.h
+++ b/esphome/components/esp32_ble/ble.h
@@ -156,8 +156,8 @@ class ESP32BLE : public Component {
 #endif
   static void gap_event_handler(esp_gap_ble_cb_event_t event, esp_ble_gap_cb_param_t *param);
 
-  // Handle non-ACTIVE state transitions (DISABLE, ENABLE, OFF, DISABLED).
-  // Extracted from loop() to keep the hot event-processing path small.
+  // Handle DISABLE and ENABLE transitions when not in the ACTIVE state.
+  // Other non-ACTIVE states (e.g. OFF, DISABLED) are currently treated as no-ops.
   void __attribute__((noinline)) loop_handle_state_transition_not_active_();
 
   bool ble_setup_();
diff --git a/esphome/components/esp8266/preferences.cpp b/esphome/components/esp8266/preferences.cpp
index f037b881a8..e749b1f633 100644
--- a/esphome/components/esp8266/preferences.cpp
+++ b/esphome/components/esp8266/preferences.cpp
@@ -33,6 +33,10 @@ static constexpr uint32_t MAX_PREFERENCE_WORDS = 255;
 
 #define ESP_RTC_USER_MEM ((uint32_t *) ESP_RTC_USER_MEM_START)
 
+// Flash storage size depends on esp8266 -> restore_from_flash YAML option (default: false).
+// When enabled (USE_ESP8266_PREFERENCES_FLASH), all preferences default to flash and need
+// 128 words (512 bytes). When disabled, only explicit flash prefs use this storage so
+// 64 words (256 bytes) suffices since most preferences go to RTC memory instead.
 #ifdef USE_ESP8266_PREFERENCES_FLASH
 static constexpr uint32_t ESP8266_FLASH_STORAGE_SIZE = 128;
 #else
@@ -127,9 +131,11 @@ static bool load_from_rtc(size_t offset, uint32_t *data, size_t len) {
   return true;
 }
 
-// Stack buffer size - 16 words total: up to 15 words of preference data + 1 word CRC (60 bytes of preference data)
-// This handles virtually all real-world preferences without heap allocation
-static constexpr size_t PREF_BUFFER_WORDS = 16;
+// Maximum buffer for any single preference - bounded by storage sizes.
+// Flash prefs: bounded by ESP8266_FLASH_STORAGE_SIZE (128 or 64 words).
+// RTC prefs: bounded by RTC_NORMAL_REGION_WORDS (96) - a single pref can't span both RTC regions.
+static constexpr size_t PREF_MAX_BUFFER_WORDS =
+    ESP8266_FLASH_STORAGE_SIZE > RTC_NORMAL_REGION_WORDS ? ESP8266_FLASH_STORAGE_SIZE : RTC_NORMAL_REGION_WORDS;
 
 class ESP8266PreferenceBackend : public ESPPreferenceBackend {
  public:
@@ -141,15 +147,13 @@ class ESP8266PreferenceBackend : public ESPPreferenceBackend {
   bool save(const uint8_t *data, size_t len) override {
     if (bytes_to_words(len) != this->length_words)
       return false;
-
     const size_t buffer_size = static_cast<size_t>(this->length_words) + 1;
-    SmallBufferWithHeapFallback<PREF_BUFFER_WORDS, uint32_t> buffer_alloc(buffer_size);
-    uint32_t *buffer = buffer_alloc.get();
+    if (buffer_size > PREF_MAX_BUFFER_WORDS)
+      return false;
+    uint32_t buffer[PREF_MAX_BUFFER_WORDS];
     memset(buffer, 0, buffer_size * sizeof(uint32_t));
-
     memcpy(buffer, data, len);
     buffer[this->length_words] = calculate_crc(buffer, buffer + this->length_words, this->type);
-
     return this->in_flash ? save_to_flash(this->offset, buffer, buffer_size)
                           : save_to_rtc(this->offset, buffer, buffer_size);
   }
@@ -157,19 +161,16 @@ class ESP8266PreferenceBackend : public ESPPreferenceBackend {
   bool load(uint8_t *data, size_t len) override {
     if (bytes_to_words(len) != this->length_words)
       return false;
-
     const size_t buffer_size = static_cast<size_t>(this->length_words) + 1;
-    SmallBufferWithHeapFallback<PREF_BUFFER_WORDS, uint32_t> buffer_alloc(buffer_size);
-    uint32_t *buffer = buffer_alloc.get();
-
+    if (buffer_size > PREF_MAX_BUFFER_WORDS)
+      return false;
+    uint32_t buffer[PREF_MAX_BUFFER_WORDS];
     bool ret = this->in_flash ? load_from_flash(this->offset, buffer, buffer_size)
                               : load_from_rtc(this->offset, buffer, buffer_size);
     if (!ret)
       return false;
-
     if (buffer[this->length_words] != calculate_crc(buffer, buffer + this->length_words, this->type))
       return false;
-
     memcpy(data, buffer, len);
     return true;
   }
diff --git a/esphome/components/rp2040/preferences.cpp b/esphome/components/rp2040/preferences.cpp
index 172da32adc..fa72fd9a24 100644
--- a/esphome/components/rp2040/preferences.cpp
+++ b/esphome/components/rp2040/preferences.cpp
@@ -25,8 +25,8 @@ static uint8_t
     s_flash_storage[RP2040_FLASH_STORAGE_SIZE];  // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
 static bool s_flash_dirty = false;               // NOLINT(cppcoreguidelines-avoid-non-const-global-variables)
 
-// Stack buffer size for preferences - covers virtually all real-world preferences without heap allocation
-static constexpr size_t PREF_BUFFER_SIZE = 64;
+// No preference can exceed the total flash storage, so stack buffer covers all cases.
+static constexpr size_t PREF_MAX_BUFFER_SIZE = RP2040_FLASH_STORAGE_SIZE;
 
 extern "C" uint8_t _EEPROM_start;
 
@@ -46,14 +46,14 @@ class RP2040PreferenceBackend : public ESPPreferenceBackend {
 
   bool save(const uint8_t *data, size_t len) override {
     const size_t buffer_size = len + 1;
-    SmallBufferWithHeapFallback<PREF_BUFFER_SIZE> buffer_alloc(buffer_size);
-    uint8_t *buffer = buffer_alloc.get();
-
+    if (buffer_size > PREF_MAX_BUFFER_SIZE)
+      return false;
+    uint8_t buffer[PREF_MAX_BUFFER_SIZE];
     memcpy(buffer, data, len);
-    buffer[len] = calculate_crc(buffer, buffer + len, type);
+    buffer[len] = calculate_crc(buffer, buffer + len, this->type);
 
     for (size_t i = 0; i < buffer_size; i++) {
-      uint32_t j = offset + i;
+      uint32_t j = this->offset + i;
       if (j >= RP2040_FLASH_STORAGE_SIZE)
         return false;
       uint8_t v = buffer[i];
@@ -66,17 +66,18 @@ class RP2040PreferenceBackend : public ESPPreferenceBackend {
   }
   bool load(uint8_t *data, size_t len) override {
     const size_t buffer_size = len + 1;
-    SmallBufferWithHeapFallback<PREF_BUFFER_SIZE> buffer_alloc(buffer_size);
-    uint8_t *buffer = buffer_alloc.get();
+    if (buffer_size > PREF_MAX_BUFFER_SIZE)
+      return false;
+    uint8_t buffer[PREF_MAX_BUFFER_SIZE];
 
     for (size_t i = 0; i < buffer_size; i++) {
-      uint32_t j = offset + i;
+      uint32_t j = this->offset + i;
       if (j >= RP2040_FLASH_STORAGE_SIZE)
         return false;
       buffer[i] = s_flash_storage[j];
     }
 
-    uint8_t crc = calculate_crc(buffer, buffer + len, type);
+    uint8_t crc = calculate_crc(buffer, buffer + len, this->type);
     if (buffer[len] != crc) {
       return false;
     }
diff --git a/esphome/components/rtttl/rtttl.cpp b/esphome/components/rtttl/rtttl.cpp
index 5a64f37da4..6e86405b74 100644
--- a/esphome/components/rtttl/rtttl.cpp
+++ b/esphome/components/rtttl/rtttl.cpp
@@ -4,28 +4,72 @@
 #include "esphome/core/log.h"
 #include "esphome/core/progmem.h"
 
-namespace esphome {
-namespace rtttl {
+namespace esphome::rtttl {
 
 static const char *const TAG = "rtttl";
 
-static const uint32_t DOUBLE_NOTE_GAP_MS = 10;
-
 // These values can also be found as constants in the Tone library (Tone.h)
 static const uint16_t NOTES[] = {0,    262,  277,  294,  311,  330,  349,  370,  392,  415,  440,  466,  494,
                                  523,  554,  587,  622,  659,  698,  740,  784,  831,  880,  932,  988,  1047,
                                  1109, 1175, 1245, 1319, 1397, 1480, 1568, 1661, 1760, 1865, 1976, 2093, 2217,
                                  2349, 2489, 2637, 2794, 2960, 3136, 3322, 3520, 3729, 3951};
 
-static const uint16_t I2S_SPEED = 1000;
+#if defined(USE_OUTPUT) || defined(USE_SPEAKER)
+static const uint32_t DOUBLE_NOTE_GAP_MS = 10;
+#endif  // USE_OUTPUT || USE_SPEAKER
 
-#undef HALF_PI
-static const double HALF_PI = 1.5707963267948966192313216916398;
+#ifdef USE_SPEAKER
+static const size_t SAMPLE_BUFFER_SIZE = 2048;
+
+struct SpeakerSample {
+  int8_t left{0};
+  int8_t right{0};
+};
 
 inline double deg2rad(double degrees) {
   static const double PI_ON_180 = 4.0 * atan(1.0) / 180.0;
   return degrees * PI_ON_180;
 }
+#endif  // USE_SPEAKER
+
+#if ESPHOME_LOG_LEVEL >= ESPHOME_LOG_LEVEL_VERBOSE
+// RTTTL state strings indexed by State enum (0-4): STOPPED, INIT, STARTING, RUNNING, STOPPING, plus UNKNOWN fallback
+PROGMEM_STRING_TABLE(RtttlStateStrings, "State::STOPPED", "State::INIT", "State::STARTING", "State::RUNNING",
+                     "State::STOPPING", "UNKNOWN");
+
+static const LogString *state_to_string(State state) {
+  return RtttlStateStrings::get_log_str(static_cast<uint8_t>(state), RtttlStateStrings::LAST_INDEX);
+}
+#endif  // ESPHOME_LOG_LEVEL >= ESPHOME_LOG_LEVEL_VERBOSE
+
+static uint8_t note_index_from_char(char note) {
+  switch (note) {
+    case 'c':
+      return 1;
+    // 'c#': 2
+    case 'd':
+      return 3;
+    // 'd#': 4
+    case 'e':
+      return 5;
+    case 'f':
+      return 6;
+    // 'f#': 7
+    case 'g':
+      return 8;
+    // 'g#': 9
+    case 'a':
+      return 10;
+    // 'a#': 11
+    // Support both 'b' (English notation for B natural) and 'h' (German notation for B natural)
+    case 'b':
+    case 'h':
+      return 12;
+    case 'p':
+    default:
+      return 0;
+  }
+}
 
 void Rtttl::dump_config() {
   ESP_LOGCONFIG(TAG,
@@ -34,161 +78,34 @@ void Rtttl::dump_config() {
                 this->gain_);
 }
 
-void Rtttl::play(std::string rtttl) {
-  if (this->state_ != State::STATE_STOPPED && this->state_ != State::STATE_STOPPING) {
-    size_t pos = this->rtttl_.find(':');
-    size_t len = (pos != std::string::npos) ? pos : this->rtttl_.length();
-    ESP_LOGW(TAG, "Already playing: %.*s", (int) len, this->rtttl_.c_str());
-    return;
-  }
-
-  this->rtttl_ = std::move(rtttl);
-
-  this->default_duration_ = 4;
-  this->default_octave_ = 6;
-  this->note_duration_ = 0;
-
-  int bpm = 63;
-  uint8_t num;
-
-  // Get name
-  this->position_ = this->rtttl_.find(':');
-
-  // it's somewhat documented to be up to 10 characters but let's be a bit flexible here
-  if (this->position_ == std::string::npos || this->position_ > 15) {
-    ESP_LOGE(TAG, "Unable to determine name; missing ':'");
-    return;
-  }
-
-  ESP_LOGD(TAG, "Playing song %.*s", (int) this->position_, this->rtttl_.c_str());
-
-  // get default duration
-  this->position_ = this->rtttl_.find("d=", this->position_);
-  if (this->position_ == std::string::npos) {
-    ESP_LOGE(TAG, "Missing 'd='");
-    return;
-  }
-  this->position_ += 2;
-  num = this->get_integer_();
-  if (num > 0)
-    this->default_duration_ = num;
-
-  // get default octave
-  this->position_ = this->rtttl_.find("o=", this->position_);
-  if (this->position_ == std::string::npos) {
-    ESP_LOGE(TAG, "Missing 'o=");
-    return;
-  }
-  this->position_ += 2;
-  num = get_integer_();
-  if (num >= 3 && num <= 7)
-    this->default_octave_ = num;
-
-  // get BPM
-  this->position_ = this->rtttl_.find("b=", this->position_);
-  if (this->position_ == std::string::npos) {
-    ESP_LOGE(TAG, "Missing b=");
-    return;
-  }
-  this->position_ += 2;
-  num = get_integer_();
-  if (num != 0)
-    bpm = num;
-
-  this->position_ = this->rtttl_.find(':', this->position_);
-  if (this->position_ == std::string::npos) {
-    ESP_LOGE(TAG, "Missing second ':'");
-    return;
-  }
-  this->position_++;
-
-  // BPM usually expresses the number of quarter notes per minute
-  this->wholenote_ = 60 * 1000L * 4 / bpm;  // this is the time for whole note (in milliseconds)
-
-  this->output_freq_ = 0;
-  this->last_note_ = millis();
-  this->note_duration_ = 1;
-
-#ifdef USE_SPEAKER
-  if (this->speaker_ != nullptr) {
-    this->set_state_(State::STATE_INIT);
-    this->samples_sent_ = 0;
-    this->samples_count_ = 0;
-  }
-#endif
-#ifdef USE_OUTPUT
-  if (this->output_ != nullptr) {
-    this->set_state_(State::STATE_RUNNING);
-  }
-#endif
-}
-
-void Rtttl::stop() {
-#ifdef USE_OUTPUT
-  if (this->output_ != nullptr) {
-    this->output_->set_level(0.0);
-    this->set_state_(STATE_STOPPED);
-  }
-#endif
-#ifdef USE_SPEAKER
-  if (this->speaker_ != nullptr) {
-    if (this->speaker_->is_running()) {
-      this->speaker_->stop();
-    }
-    this->set_state_(STATE_STOPPING);
-  }
-#endif
-  this->position_ = this->rtttl_.length();
-  this->note_duration_ = 0;
-}
-
-void Rtttl::finish_() {
-  ESP_LOGV(TAG, "Rtttl::finish_()");
-#ifdef USE_OUTPUT
-  if (this->output_ != nullptr) {
-    this->output_->set_level(0.0);
-    this->set_state_(State::STATE_STOPPED);
-  }
-#endif
-#ifdef USE_SPEAKER
-  if (this->speaker_ != nullptr) {
-    SpeakerSample sample[2];
-    sample[0].left = 0;
-    sample[0].right = 0;
-    sample[1].left = 0;
-    sample[1].right = 0;
-    this->speaker_->play((uint8_t *) (&sample), 8);
-    this->speaker_->finish();
-    this->set_state_(State::STATE_STOPPING);
-  }
-#endif
-  // Ensure no more notes are played in case finish_() is called for an error.
-  this->position_ = this->rtttl_.length();
-  this->note_duration_ = 0;
-}
-
 void Rtttl::loop() {
-  if (this->state_ == State::STATE_STOPPED) {
+  if (this->state_ == State::STOPPED) {
     this->disable_loop();
     return;
   }
 
+#ifdef USE_OUTPUT
+  if (this->output_ != nullptr && millis() - this->last_note_ < this->note_duration_) {
+    return;
+  }
+#endif  // USE_OUTPUT
+
 #ifdef USE_SPEAKER
   if (this->speaker_ != nullptr) {
-    if (this->state_ == State::STATE_STOPPING) {
+    if (this->state_ == State::STOPPING) {
       if (this->speaker_->is_stopped()) {
-        this->set_state_(State::STATE_STOPPED);
+        this->set_state_(State::STOPPED);
       } else {
         return;
       }
-    } else if (this->state_ == State::STATE_INIT) {
+    } else if (this->state_ == State::INIT) {
       if (this->speaker_->is_stopped()) {
         this->speaker_->start();
-        this->set_state_(State::STATE_STARTING);
+        this->set_state_(State::STARTING);
       }
-    } else if (this->state_ == State::STATE_STARTING) {
+    } else if (this->state_ == State::STARTING) {
       if (this->speaker_->is_running()) {
-        this->set_state_(State::STATE_RUNNING);
+        this->set_state_(State::RUNNING);
       }
     }
     if (!this->speaker_->is_running()) {
@@ -230,19 +147,17 @@ void Rtttl::loop() {
       }
     }
   }
-#endif
-#ifdef USE_OUTPUT
-  if (this->output_ != nullptr && millis() - this->last_note_ < this->note_duration_)
-    return;
-#endif
+#endif  // USE_SPEAKER
+
   if (this->position_ >= this->rtttl_.length()) {
     this->finish_();
     return;
   }
 
   // align to note: most rtttl's out there does not add and space after the ',' separator but just in case...
-  while (this->rtttl_[this->position_] == ',' || this->rtttl_[this->position_] == ' ')
+  while (this->rtttl_[this->position_] == ',' || this->rtttl_[this->position_] == ' ') {
     this->position_++;
+  }
 
   // first, get note duration, if available
   uint8_t num = this->get_integer_();
@@ -254,35 +169,8 @@ void Rtttl::loop() {
         this->wholenote_ / this->default_duration_;  // we will need to check if we are a dotted note after
   }
 
-  uint8_t note;
+  uint8_t note = note_index_from_char(this->rtttl_[this->position_]);
 
-  switch (this->rtttl_[this->position_]) {
-    case 'c':
-      note = 1;
-      break;
-    case 'd':
-      note = 3;
-      break;
-    case 'e':
-      note = 5;
-      break;
-    case 'f':
-      note = 6;
-      break;
-    case 'g':
-      note = 8;
-      break;
-    case 'a':
-      note = 10;
-      break;
-    case 'h':
-    case 'b':
-      note = 12;
-      break;
-    case 'p':
-    default:
-      note = 0;
-  }
   this->position_++;
 
   // now, get optional '#' sharp
@@ -292,7 +180,7 @@ void Rtttl::loop() {
   }
 
   // now, get scale
-  uint8_t scale = get_integer_();
+  uint8_t scale = this->get_integer_();
   if (scale == 0) {
     scale = this->default_octave_;
   }
@@ -345,7 +233,8 @@ void Rtttl::loop() {
       this->output_->set_level(0.0);
     }
   }
-#endif
+#endif  // USE_OUTPUT
+
 #ifdef USE_SPEAKER
   if (this->speaker_ != nullptr) {
     this->samples_sent_ = 0;
@@ -370,20 +259,152 @@ void Rtttl::loop() {
     }
     // Convert from frequency in Hz to high and low samples in fixed point
   }
-#endif
+#endif  // USE_SPEAKER
 
   this->last_note_ = millis();
 }
 
-#if ESPHOME_LOG_LEVEL >= ESPHOME_LOG_LEVEL_VERBOSE
-// RTTTL state strings indexed by State enum (0-4): STOPPED, INIT, STARTING, RUNNING, STOPPING, plus UNKNOWN fallback
-PROGMEM_STRING_TABLE(RtttlStateStrings, "STATE_STOPPED", "STATE_INIT", "STATE_STARTING", "STATE_RUNNING",
-                     "STATE_STOPPING", "UNKNOWN");
+void Rtttl::play(std::string rtttl) {
+  if (this->state_ != State::STOPPED && this->state_ != State::STOPPING) {
+    size_t pos = this->rtttl_.find(':');
+    size_t len = (pos != std::string::npos) ? pos : this->rtttl_.length();
+    ESP_LOGW(TAG, "Already playing: %.*s", (int) len, this->rtttl_.c_str());
+    return;
+  }
 
-static const LogString *state_to_string(State state) {
-  return RtttlStateStrings::get_log_str(static_cast<uint8_t>(state), RtttlStateStrings::LAST_INDEX);
+  this->rtttl_ = std::move(rtttl);
+
+  this->default_duration_ = 4;
+  this->default_octave_ = 6;
+  this->note_duration_ = 0;
+
+  int bpm = 63;
+  uint8_t num;
+
+  // Get name
+  this->position_ = this->rtttl_.find(':');
+
+  // it's somewhat documented to be up to 10 characters but let's be a bit flexible here
+  if (this->position_ == std::string::npos || this->position_ > 15) {
+    ESP_LOGE(TAG, "Unable to determine name; missing ':'");
+    return;
+  }
+
+  ESP_LOGD(TAG, "Playing song %.*s", (int) this->position_, this->rtttl_.c_str());
+
+  // get default duration
+  this->position_ = this->rtttl_.find("d=", this->position_);
+  if (this->position_ == std::string::npos) {
+    ESP_LOGE(TAG, "Missing 'd='");
+    return;
+  }
+  this->position_ += 2;
+  num = this->get_integer_();
+  if (num > 0) {
+    this->default_duration_ = num;
+  }
+
+  // get default octave
+  this->position_ = this->rtttl_.find("o=", this->position_);
+  if (this->position_ == std::string::npos) {
+    ESP_LOGE(TAG, "Missing 'o=");
+    return;
+  }
+  this->position_ += 2;
+  num = this->get_integer_();
+  if (num >= 3 && num <= 7) {
+    this->default_octave_ = num;
+  }
+
+  // get BPM
+  this->position_ = this->rtttl_.find("b=", this->position_);
+  if (this->position_ == std::string::npos) {
+    ESP_LOGE(TAG, "Missing b=");
+    return;
+  }
+  this->position_ += 2;
+  num = this->get_integer_();
+  if (num != 0) {
+    bpm = num;
+  }
+
+  this->position_ = this->rtttl_.find(':', this->position_);
+  if (this->position_ == std::string::npos) {
+    ESP_LOGE(TAG, "Missing second ':'");
+    return;
+  }
+  this->position_++;
+
+  // BPM usually expresses the number of quarter notes per minute
+  this->wholenote_ = 60 * 1000L * 4 / bpm;  // this is the time for whole note (in milliseconds)
+
+  this->output_freq_ = 0;
+  this->last_note_ = millis();
+  this->note_duration_ = 1;
+
+#ifdef USE_OUTPUT
+  if (this->output_ != nullptr) {
+    this->set_state_(State::RUNNING);
+  }
+#endif  // USE_OUTPUT
+
+#ifdef USE_SPEAKER
+  if (this->speaker_ != nullptr) {
+    this->set_state_(State::INIT);
+    this->samples_sent_ = 0;
+    this->samples_count_ = 0;
+  }
+#endif  // USE_SPEAKER
+}
+
+void Rtttl::stop() {
+#ifdef USE_OUTPUT
+  if (this->output_ != nullptr) {
+    this->output_->set_level(0.0);
+    this->set_state_(State::STOPPED);
+  }
+#endif  // USE_OUTPUT
+
+#ifdef USE_SPEAKER
+  if (this->speaker_ != nullptr) {
+    if (this->speaker_->is_running()) {
+      this->speaker_->stop();
+    }
+    this->set_state_(State::STOPPING);
+  }
+#endif  // USE_SPEAKER
+
+  this->position_ = this->rtttl_.length();
+  this->note_duration_ = 0;
+}
+
+void Rtttl::finish_() {
+  ESP_LOGV(TAG, "Rtttl::finish_()");
+
+#ifdef USE_OUTPUT
+  if (this->output_ != nullptr) {
+    this->output_->set_level(0.0);
+    this->set_state_(State::STOPPED);
+  }
+#endif  // USE_OUTPUT
+
+#ifdef USE_SPEAKER
+  if (this->speaker_ != nullptr) {
+    SpeakerSample sample[2];
+    sample[0].left = 0;
+    sample[0].right = 0;
+    sample[1].left = 0;
+    sample[1].right = 0;
+    this->speaker_->play((uint8_t *) (&sample), 8);
+    this->speaker_->finish();
+    this->set_state_(State::STOPPING);
+  }
+#endif  // USE_SPEAKER
+
+  // Ensure no more notes are played in case finish_() is called for an error.
+  this->position_ = this->rtttl_.length();
+  this->note_duration_ = 0;
 }
-#endif
 
 void Rtttl::set_state_(State state) {
   State old_state = this->state_;
@@ -391,15 +412,14 @@ void Rtttl::set_state_(State state) {
   ESP_LOGV(TAG, "State changed from %s to %s", LOG_STR_ARG(state_to_string(old_state)),
            LOG_STR_ARG(state_to_string(state)));
 
-  // Clear loop_done when transitioning from STOPPED to any other state
-  if (state == State::STATE_STOPPED) {
+  // Clear loop_done when transitioning from `State::STOPPED` to any other state
+  if (state == State::STOPPED) {
     this->disable_loop();
     this->on_finished_playback_callback_.call();
     ESP_LOGD(TAG, "Playback finished");
-  } else if (old_state == State::STATE_STOPPED) {
+  } else if (old_state == State::STOPPED) {
     this->enable_loop();
   }
 }
 
-}  // namespace rtttl
-}  // namespace esphome
+}  // namespace esphome::rtttl
diff --git a/esphome/components/rtttl/rtttl.h b/esphome/components/rtttl/rtttl.h
index 1e924a897c..6f5df07766 100644
--- a/esphome/components/rtttl/rtttl.h
+++ b/esphome/components/rtttl/rtttl.h
@@ -5,48 +5,41 @@
 
 #ifdef USE_OUTPUT
 #include "esphome/components/output/float_output.h"
-#endif
+#endif  // USE_OUTPUT
 
 #ifdef USE_SPEAKER
 #include "esphome/components/speaker/speaker.h"
-#endif
+#endif  // USE_SPEAKER
 
-namespace esphome {
-namespace rtttl {
+namespace esphome::rtttl {
 
-enum State : uint8_t {
-  STATE_STOPPED = 0,
-  STATE_INIT,
-  STATE_STARTING,
-  STATE_RUNNING,
-  STATE_STOPPING,
+enum class State : uint8_t {
+  STOPPED = 0,
+  INIT,
+  STARTING,
+  RUNNING,
+  STOPPING,
 };
 
-#ifdef USE_SPEAKER
-static const size_t SAMPLE_BUFFER_SIZE = 2048;
-
-struct SpeakerSample {
-  int8_t left{0};
-  int8_t right{0};
-};
-#endif
-
 class Rtttl : public Component {
  public:
 #ifdef USE_OUTPUT
   void set_output(output::FloatOutput *output) { this->output_ = output; }
-#endif
+#endif  // USE_OUTPUT
+
 #ifdef USE_SPEAKER
   void set_speaker(speaker::Speaker *speaker) { this->speaker_ = speaker; }
-#endif
-  float get_gain() { return gain_; }
-  void set_gain(float gain) { this->gain_ = clamp(gain, 0.0f, 1.0f); }
+#endif  // USE_SPEAKER
+
+  void dump_config() override;
+  void loop() override;
   void play(std::string rtttl);
   void stop();
-  void dump_config() override;
 
-  bool is_playing() { return this->state_ != State::STATE_STOPPED; }
-  void loop() override;
+  float get_gain() { return this->gain_; }
+  void set_gain(float gain) { this->gain_ = clamp(gain, 0.0f, 1.0f); }
+
+  bool is_playing() { return this->state_ != State::STOPPED; }
 
   void add_on_finished_playback_callback(std::function<void()> callback) {
     this->on_finished_playback_callback_.add(std::move(callback));
@@ -90,12 +83,12 @@ class Rtttl : public Component {
   /// The gain of the output.
   float gain_{0.6f};
   /// The current state of the RTTTL player.
-  State state_{State::STATE_STOPPED};
+  State state_{State::STOPPED};
 
 #ifdef USE_OUTPUT
   /// The output to write the sound to.
   output::FloatOutput *output_;
-#endif
+#endif  // USE_OUTPUT
 
 #ifdef USE_SPEAKER
   /// The speaker to write the sound to.
@@ -110,8 +103,7 @@ class Rtttl : public Component {
   int samples_count_{0};
   /// The number of samples for the gap between notes.
   int samples_gap_{0};
-
-#endif
+#endif  // USE_SPEAKER
 
   /// The callback to call when playback is finished.
   CallbackManager<void()> on_finished_playback_callback_;
@@ -145,5 +137,4 @@ class FinishedPlaybackTrigger : public Trigger<> {
   }
 };
 
-}  // namespace rtttl
-}  // namespace esphome
+}  // namespace esphome::rtttl