[i2s_audio] Fix mono sample swap and block 8-bit mono on ESP32 (#14516)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: J. Nick Koston <nick@koston.org>

esphome/components/i2s_audio/microphone/__init__.py

@@ -46,6 +46,12 @@ def _validate_esp32_variant(config):
     if config[CONF_ADC_TYPE] == "external":
         if config[CONF_PDM] and variant not in PDM_VARIANTS:
             raise cv.Invalid(f"{variant} does not support PDM")
+        if (
+            variant == esp32.VARIANT_ESP32
+            and config.get(CONF_BITS_PER_SAMPLE) == 8
+            and config.get(CONF_CHANNEL) in (CONF_LEFT, CONF_RIGHT)
+        ):
+            raise cv.Invalid("8-bit mono mode is not supported on ESP32")
         return config
     if config[CONF_ADC_TYPE] == "internal":
         if variant not in INTERNAL_ADC_VARIANTS:

esphome/components/i2s_audio/microphone/i2s_audio_microphone.cpp

@@ -281,7 +281,7 @@ bool I2SAudioMicrophone::start_driver_() {
   }
 
   /* Before reading data, start the RX channel first */
-  i2s_channel_enable(this->rx_handle_);
+  err = i2s_channel_enable(this->rx_handle_);
   if (err != ESP_OK) {
     ESP_LOGE(TAG, "Enabling failed: %s", esp_err_to_name(err));
     return false;
@@ -454,13 +454,14 @@ size_t I2SAudioMicrophone::read_(uint8_t *buf, size_t len, TickType_t ticks_to_w
   }
   this->status_clear_warning();
 #if defined(USE_ESP32_VARIANT_ESP32) and not defined(USE_I2S_LEGACY)
-  // For ESP32 8/16 bit standard mono mode samples need to be switched.
-  if (this->slot_mode_ == I2S_SLOT_MODE_MONO && this->slot_bit_width_ <= 16 && !this->pdm_) {
-    size_t samples_read = bytes_read / sizeof(int16_t);
-    for (int i = 0; i < samples_read; i += 2) {
-      int16_t tmp = buf[i];
-      buf[i] = buf[i + 1];
-      buf[i + 1] = tmp;
+  // For ESP32 16-bit standard mono mode, adjacent samples need to be swapped.
+  if (this->slot_mode_ == I2S_SLOT_MODE_MONO && this->slot_bit_width_ == I2S_SLOT_BIT_WIDTH_16BIT && !this->pdm_) {
+    int16_t *samples = reinterpret_cast<int16_t *>(buf);
+    size_t sample_count = bytes_read / sizeof(int16_t);
+    for (size_t i = 0; i + 1 < sample_count; i += 2) {
+      int16_t tmp = samples[i];
+      samples[i] = samples[i + 1];
+      samples[i + 1] = tmp;
     }
   }
 #endif

esphome/components/i2s_audio/speaker/__init__.py

@@ -100,11 +100,16 @@ def _set_stream_limits(config):
 
 
 def _validate_esp32_variant(config):
-    if config[CONF_DAC_TYPE] != "internal":
-        return config
     variant = esp32.get_esp32_variant()
-    if variant not in INTERNAL_DAC_VARIANTS:
-        raise cv.Invalid(f"{variant} does not have an internal DAC")
+    if config[CONF_DAC_TYPE] == "internal":
+        if variant not in INTERNAL_DAC_VARIANTS:
+            raise cv.Invalid(f"{variant} does not have an internal DAC")
+    elif (
+        variant == esp32.VARIANT_ESP32
+        and config.get(CONF_BITS_PER_SAMPLE) == 8
+        and config.get(CONF_CHANNEL) in (CONF_MONO, CONF_LEFT, CONF_RIGHT)
+    ):
+        raise cv.Invalid("8-bit mono mode is not supported on ESP32")
     return config
 
 

esphome/components/i2s_audio/speaker/i2s_audio_speaker.cpp

@@ -372,15 +372,15 @@ void I2SAudioSpeaker::speaker_task(void *params) {
         }
 
 #ifdef USE_ESP32_VARIANT_ESP32
-        // For ESP32 8/16 bit mono mode samples need to be switched.
+        // For ESP32 16-bit mono mode, adjacent samples need to be swapped.
         if (this_speaker->current_stream_info_.get_channels() == 1 &&
-            this_speaker->current_stream_info_.get_bits_per_sample() <= 16) {
-          size_t len = bytes_read / sizeof(int16_t);
-          int16_t *tmp_buf = (int16_t *) new_data;
-          for (size_t i = 0; i < len; i += 2) {
-            int16_t tmp = tmp_buf[i];
-            tmp_buf[i] = tmp_buf[i + 1];
-            tmp_buf[i + 1] = tmp;
+            this_speaker->current_stream_info_.get_bits_per_sample() == 16) {
+          int16_t *samples = reinterpret_cast<int16_t *>(new_data);
+          size_t sample_count = bytes_read / sizeof(int16_t);
+          for (size_t i = 0; i + 1 < sample_count; i += 2) {
+            int16_t tmp = samples[i];
+            samples[i] = samples[i + 1];
+            samples[i + 1] = tmp;
           }
         }
 #endif

esphome/components/speaker/media_player/audio_pipeline.cpp

@@ -504,7 +504,7 @@ void AudioPipeline::decode_task(void *params) {
         if (!started_playback && has_stream_info) {
           // Verify enough data is available before starting playback
           std::shared_ptr<RingBuffer> temp_ring_buffer = this_pipeline->raw_file_ring_buffer_.lock();
-          if (temp_ring_buffer->available() >= initial_bytes_to_buffer) {
+          if (temp_ring_buffer != nullptr && temp_ring_buffer->available() >= initial_bytes_to_buffer) {
             started_playback = true;
           }
         }