fix: always initialize face_to_paste_pil
This commit is contained in:
+1
-2
@@ -93,8 +93,7 @@ class VideoInputPreprocessor:
|
||||
print("[Warning] No landmarks found in source 'face_rgba' image. Angle matching disabled.")
|
||||
match_angle_and_size = False
|
||||
|
||||
if not match_angle_and_size:
|
||||
face_to_paste_pil = tensor_to_pil(face_rgba[0], mode='RGBA')
|
||||
face_to_paste_pil = tensor_to_pil(face_rgba[0], mode='RGBA')
|
||||
|
||||
processed_frames_tensors = []
|
||||
|
||||
|
||||
Reference in New Issue
Block a user