diff --git a/codes/data/audio/nv_tacotron_dataset.py b/codes/data/audio/nv_tacotron_dataset.py
index f7a3f45d..5099cdee 100644
--- a/codes/data/audio/nv_tacotron_dataset.py
+++ b/codes/data/audio/nv_tacotron_dataset.py
@@ -33,7 +33,7 @@ def load_voxpopuli(filename):
                 continue
             file, raw_text, norm_text, speaker_id, split, gender = line
             year = file[:4]
-            filepaths_and_text.append([os.path.join(base, year, file), raw_text])
+            filepaths_and_text.append([os.path.join(base, year, f'{file}.ogg'), raw_text])
     return filepaths_and_text