Update app.py

app.py

@@ -450,23 +450,25 @@ def process(
     if t2v:
         default_height, default_width = 640, 640
         input_image = np.ones((default_height, default_width, 3), dtype=np.uint8) * 255
-        print("No input image provided. Using a blank white image.")
+        print("Using blank white image for text-to-video mode")
     else:
-        composite_rgba_uint8 = input_image["composite"]
-
-        rgb_uint8 = composite_rgba_uint8[:, :, :3]
-        mask_uint8 = composite_rgba_uint8[:, :, 3]
-
-        h, w = rgb_uint8.shape[:2]
-        background_uint8 = np.full((h, w, 3), 255, dtype=np.uint8)
-
-        alpha_normalized_float32 = mask_uint8.astype(np.float32) / 255.0
-        alpha_mask_float32 = np.stack([alpha_normalized_float32]*3, axis=2)
-
-        blended_image_float32 = rgb_uint8.astype(np.float32) * alpha_mask_float32 + \
+        if isinstance(input_image, dict) and "composite" in input_image:
+            # Handle uploaded image with alpha channel
+            composite_rgba_uint8 = input_image["composite"]
+            rgb_uint8 = composite_rgba_uint8[:, :, :3]
+            mask_uint8 = composite_rgba_uint8[:, :, 3]
+            h, w = rgb_uint8.shape[:2]
+            background_uint8 = np.full((h, w, 3), 255, dtype=np.uint8)
+            alpha_normalized_float32 = mask_uint8.astype(np.float32) / 255.0
+            alpha_mask_float32 = np.stack([alpha_normalized_float32]*3, axis=2)
+            blended_image_float32 = rgb_uint8.astype(np.float32) * alpha_mask_float32 + \
                                 background_uint8.astype(np.float32) * (1.0 - alpha_mask_float32)
-
-        input_image = np.clip(blended_image_float32, 0, 255).astype(np.uint8)
+            input_image = np.clip(blended_image_float32, 0, 255).astype(np.uint8)
+        elif input_image is None:
+            raise ValueError("Please provide an input image or enable Text to Video mode")
+        else:
+            # Handle regular RGB image
+            input_image = input_image.astype(np.uint8)
 
     yield None, None, '', '', gr.update(interactive=False), gr.update(interactive=True)