update description

app.py

@@ -240,7 +240,7 @@ def process_video(video_path, prompt, target_fps=30, max_dimension=640):
         print(traceback.format_exc())  # This will print the full stack trace
         return None
     
-@spaces.GPU(duration=600)
+@spaces.GPU(duration=300)
 def segment_video(video_file, prompt):
     if video_file is None:
         return None
@@ -258,7 +258,7 @@ demo = gr.Interface(
     description="""
     This demo uses [Florence-2](https://huggingface.co/microsoft/Florence-2-large), to enable text-prompted object detection for [SAM2](https://github.com/facebookresearch/segment-anything).
 
-    1. Upload a short video (< 10 sec , you can fork this space on larger GPU for longer vids) 
+    1. Upload a short video (< 6-7 seconds , you can clone this space on larger GPU for longer vids) 
     2. Describe the object to segment (The object should be visible in the first frame).
     3. Get your segmented video.
     """