diff --git a/advanced_training_notebook.ipynb b/advanced_training_notebook.ipynb
index 66c9c92..51c3046 100644
--- a/advanced_training_notebook.ipynb
+++ b/advanced_training_notebook.ipynb
@@ -136,7 +136,7 @@
     "id": "YJRG4Qvo9nXG"
    },
    "outputs": [],
-    "source": [
+   "source": [
     "# Downloads audio data for augmentation. This can be slow!\n",
     "# Borrowed from openWakeWord's automatic_model_training.ipynb, accessed March 4, 2024\n",
     "#\n",
@@ -205,8 +205,8 @@
     "    audioset_dataset = audioset_dataset.cast_column(\"audio\", Audio(sampling_rate=16000))\n",
     "\n",
     "    corrupted_files = []\n",
-    "    print(\"Converting Audioset files to 16kHz WAV...\")\n",
     "\n",
+    "    print(\"Converting Audioset files to 16kHz WAV...\")\n",
     "    for row in tqdm(audioset_dataset, desc=\"Processing Audioset files\"):\n",
     "        try:\n",
     "            # Define output file name and path\n",
@@ -215,6 +215,8 @@
     "            \n",
     "            # Read and convert audio\n",
     "            data = row[\"audio\"][\"array\"]\n",
+    "            if data is None or len(data) == 0:\n",
+    "                raise ValueError(f\"Empty audio data in file: {row['audio']['path']}\")\n",
     "            scipy.io.wavfile.write(output_path, 16000, (data * 32767).astype(np.int16))\n",
     "        \n",
     "        except Exception as e:\n",