diff --git a/headmic.py b/headmic.py
index b41896c..87e01f6 100644
--- a/headmic.py
+++ b/headmic.py
@@ -465,15 +465,19 @@ async def startup():
 
     # --- Sound classifier (optional) ---
     model_dir = Path(__file__).parent / "models"
+    edgetpu_model_path = model_dir / "yamnet_edgetpu.tflite"
     model_path = model_dir / "yamnet.tflite"
     class_map_path = model_dir / "yamnet_class_map.csv"
-    if model_path.exists() and class_map_path.exists():
+    # Prefer Edge TPU model if available
+    use_edgetpu = edgetpu_model_path.exists()
+    active_model = edgetpu_model_path if use_edgetpu else model_path
+    if active_model.exists() and class_map_path.exists():
         try:
             from sound_id import SoundClassifier
-            sound_classifier = SoundClassifier(str(model_path), str(class_map_path))
+            sound_classifier = SoundClassifier(str(active_model), str(class_map_path), use_edgetpu=use_edgetpu)
             sound_ring_buffer = collections.deque(maxlen=31)
             state.sound_classification_enabled = True
-            logger.info("Sound classification enabled (YAMNet)")
+            logger.info("Sound classification enabled (YAMNet %s)", "Edge TPU" if use_edgetpu else "CPU")
 
             sc_thread = threading.Thread(target=sound_classifier_loop, daemon=True)
             sc_thread.start()