Update app.py

app.py

@@ -30,13 +30,26 @@ def load_cached_example_outputs(example_index: int) -> Tuple[str, str]:
     cached_dir = os.path.join(CACHE_DIR, str(example_index))  # Use the example index to find the directory
     cached_image_path = os.path.join(cached_dir, "processed_image.png")
     cached_audio_path = os.path.join(cached_dir, "audio.wav")
+    cached_video_path = os.path.join(cached_dir, "processed_video.mp4")  # Support video output too
 
-    # Ensure cached files exist
+    # Ensure cached files exist (either image or video)
     if os.path.exists(cached_image_path) and os.path.exists(cached_audio_path):
         return cached_image_path, cached_audio_path
+    elif os.path.exists(cached_video_path) and os.path.exists(cached_audio_path):
+        return cached_video_path, cached_audio_path
     else:
         raise FileNotFoundError(f"Cached outputs not found for example {example_index}")
 
+
+# detect whether the input is an image or a video
+def determine_input_type(filepath: str):
+    if filepath.endswith((".png", ".jpg", ".jpeg")):
+        return gr.Image
+    elif filepath.endswith(".mp4"):
+        return gr.Video
+    else:
+        raise ValueError("Unsupported file type")
+
 # Function to handle the example click, it now accepts arbitrary arguments
 def on_example_click(*args, **kwargs):
     return load_cached_example_outputs(1)  # Always load example 1 for now
@@ -109,12 +122,19 @@ with gr.Blocks(css=css) as demo:
                 ),
             )
 
+    def load_example_by_input_type(file_input, *args, example_index):
+        input_type = determine_input_type(file_input)
+        if input_type == gr.Video:
+            return gr.Video(label="Video Example"), on_example_click(example_index=example_index)
+        else:
+            return gr.Image(label="Image Example"), on_example_click(example_index=example_index)
+
     gr.Examples(
-        examples=[["examples/1.png", 3, "A scenic mountain view", 500]],  # Example input
-        inputs=[image, num_audios, prompt, steps],
-        outputs=[processed_image, generated_audio],
-        cache_examples=True,  # Cache examples to avoid running the model
-        fn=on_example_click  # Load the cached output when the example is clicked
+        examples=example_inputs,
+        inputs=[gr.File(label="Select image or video")],
+        outputs=[processed_media, generated_audio],
+        cache_examples=True,
+        fn=lambda *args, **kwargs: load_example_by_input_type(args[0], example_index=args[0][0]),  # Dynamically detect and process
     )
 
     gr.on(