diff --git a/color_calibration/requirements.txt b/color_calibration/requirements.txt
index 711d080..a305b5a 100644
--- a/color_calibration/requirements.txt
+++ b/color_calibration/requirements.txt
@@ -1,2 +1,2 @@
-numpy==1.20.2
+numpy==1.22.0
 tensorflow==2.4.1
diff --git a/hti/server/api/stacking_api.py b/hti/server/api/stacking_api.py
index 9dbfda2..3cd3581 100644
--- a/hti/server/api/stacking_api.py
+++ b/hti/server/api/stacking_api.py
@@ -51,7 +51,7 @@ def post(self):
         Frame.interpolate_angles(frames)
 
         image_stacker = get_image_stacker()
-        image_stacker.light_frames = frames[:10]
+        image_stacker.light_frames = frames[:10]  # TODO use all frames, provide a range option
         image_stacker.stack_image()
 
         # Update hash in app state so FE can react by reloading the preview
@@ -69,7 +69,10 @@ class StackedImagePreviewApi(Resource):
     def get(self, stacked_image_hash):
         image_stacker = get_image_stacker()
         if image_stacker.get_stacked_image_hash() == stacked_image_hash:
-            image_data_png = image_stacker.get_stacked_image(format='png')
+            image_data_png = image_stacker.get_stacked_image(
+                format='png',
+                scale_factor=2
+            )
             return send_file(image_data_png, mimetype='image/png')
 
         # TODO hash is outdated - maybe update it with whatever is new?
diff --git a/hti/server/stacking/image_stacker.py b/hti/server/stacking/image_stacker.py
index 59fd58a..b5a217d 100644
--- a/hti/server/stacking/image_stacker.py
+++ b/hti/server/stacking/image_stacker.py
@@ -27,14 +27,17 @@ def __init__(self):
 	def get_stacked_image_hash(self) -> str:
 		"""
 		Get a hash of all the parameters that went into stacked_image
-		so we can be smart about re-stacking only if settings change.
+		so we can be smart about re-stacking and fetching the updated
+		result only if any settings changed.
 		"""
 		# TODO crazy stupid hashing
 		return str(len(self.light_frames))
 
 	@staticmethod
 	def normalize_image(image):
-
+		"""
+		Fit dynamic range found in the given image into [0;1].
+		"""
 		normalized_image = np.zeros(image.shape, dtype=image.dtype)
 
 		for channel in range(image.shape[2]):