diff --git a/src/spikeinterface/postprocessing/template_similarity.py b/src/spikeinterface/postprocessing/template_similarity.py
index 27214f32e6..0e70b1f494 100644
--- a/src/spikeinterface/postprocessing/template_similarity.py
+++ b/src/spikeinterface/postprocessing/template_similarity.py
@@ -150,7 +150,6 @@ def _get_data(self):
 def compute_similarity_with_templates_array(
     templates_array, other_templates_array, method, support="union", num_shifts=0, sparsity=None, other_sparsity=None
 ):
-
     import sklearn.metrics.pairwise
 
     if method == "cosine_similarity":
@@ -223,15 +222,17 @@ def compute_similarity_with_templates_array(
                 if method == "l1":
                     norm_i = np.sum(np.abs(src))
                     norm_j = np.sum(np.abs(tgt))
-                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(src, tgt, metric="l1")
+                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(src, tgt, metric="l1").item()
                     distances[count, i, j] /= norm_i + norm_j
                 elif method == "l2":
                     norm_i = np.linalg.norm(src, ord=2)
                     norm_j = np.linalg.norm(tgt, ord=2)
-                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(src, tgt, metric="l2")
+                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(src, tgt, metric="l2").item()
                     distances[count, i, j] /= norm_i + norm_j
                 else:
-                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(src, tgt, metric="cosine")
+                    distances[count, i, j] = sklearn.metrics.pairwise.pairwise_distances(
+                        src, tgt, metric="cosine"
+                    ).item()
 
                 if same_array:
                     distances[count, j, i] = distances[count, i, j]