diff --git a/scripts/speech_recognition/convert_hf_dataset_to_nemo.py b/scripts/speech_recognition/convert_hf_dataset_to_nemo.py
index 8445a8f11754..aa29bcc1620b 100644
--- a/scripts/speech_recognition/convert_hf_dataset_to_nemo.py
+++ b/scripts/speech_recognition/convert_hf_dataset_to_nemo.py
@@ -208,7 +208,7 @@ def map_dataset_to_nemo(batch):
 
         soundfile.write(audio_filepath, batch['audio']['array'], samplerate=cfg.sampling_rate, format='wav')
 
-        batch['duration'] = librosa.get_duration(batch['audio']['array'], sr=batch['audio']['sampling_rate'])
+        batch['duration'] = librosa.get_duration(y=batch['audio']['array'], sr=batch['audio']['sampling_rate'])
         return batch
 
     return map_dataset_to_nemo