diff --git a/src/separation/convtasnet_wrapper.py b/src/separation/convtasnet_wrapper.py
index 0c08b2a..a277762 100644
--- a/src/separation/convtasnet_wrapper.py
+++ b/src/separation/convtasnet_wrapper.py
@@ -1,13 +1,13 @@
 import torch
 from asteroid.models import ConvTasNet
 
-def separate(audio, model_name='mpariente/ConvTasNet_WHAMR_enhsingle'):
+def separate(audio, model_name='mpariente/ConvTasNet_WHAM!_sepclean'):
     """
     Separates audio into sources using a pretrained Asteroid model.
 
     Args:
         audio (numpy.ndarray): The audio time series (1D numpy array).
-        model_name (str): Name of the pretrained model from Asteroid. Default is 'mpariente/ConvTasNet_WHAMR_enhsingle'.
+        model_name (str): Name of the pretrained model from Asteroid. Default is 'mpariente/ConvTasNet_WHAM!_sepclean'.
 
     Returns:
         list: List of separated sources as numpy arrays.