Update app.py

app.py

@@ -216,6 +216,15 @@ def infer_inp(prompt, audio_path, spec_with_mask, progress=gr.Progress(track_tqd
 
     return "output.wav"
 
+def create_transparent_layer(image_path):
+    """Creates a transparent PNG with the same size as the background image."""
+    background = Image.open(image_path)
+    transparent_layer = Image.new("RGBA", background.size, (0, 0, 0, 0))
+    
+    layer_path = "layer_one.png"
+    transparent_layer.save(layer_path)
+    return layer_path
+
 def load_spec_for_manual_masking(audio_path):
     # Loading
     audio, sampling_rate = load_wav(audio_path)
@@ -228,7 +237,16 @@ def load_spec_for_manual_masking(audio_path):
 
     input_spec_image_path = "input_spectrogram.png"
     raw_image.save(input_spec_image_path)
-    return input_spec_image_path
+    
+    # Create transparent layer
+    layer_one_path = create_transparent_layer(input_spec_image_path)
+    
+    # Return as EditorValue
+    return {
+        "background": input_spec_image_path,
+        "layers": [layer_one_path],
+        "composite": None
+    }
     
 
 css="""