Update app.py

app.py

@@ -17,7 +17,10 @@ zephyr_model = "HuggingFaceH4/zephyr-7b-beta"
 pipe = pipeline("text-generation", model=zephyr_model, torch_dtype=torch.bfloat16, device_map="auto")
 
 standard_sys = f"""
-
+You will be provided a list of visual events, and an audio description. All these informations come from a single video. 
+List of visual events are actually images extracted from this video every 12 frames.
+Audio events are actually the description from the audio of the video.
+Your job is to use these information to provide a short resume about what is happening in the video.
 """
 
 def extract_frames(video_in, interval=24, output_format='.jpg'):
@@ -83,7 +86,7 @@ def extract_audio(video_path):
     return "output_audio.mp3"
 
 def get_salmonn(audio_in):
-    salmonn_prompt = "Describe sound."
+    salmonn_prompt = "Please describe the audio"
     client = Client("fffiloni/SALMONN-7B-gradio")
     result = client.predict(
     		audio_in,	# filepath  in 'Audio' Audio component
@@ -141,9 +144,9 @@ def infer(video_in):
     print(formatted_captions)
 
     # Send formatted captions to LLM
-    #video_description_from_llm = llm_process(formatted_captions)
+    video_description_from_llm = llm_process(formatted_captions)
     
-    return formatted_captions
+    return video_description_from_llm
 
 with gr.Blocks() as demo :
     with gr.Column(elem_id="col-container"):