Update app.py

app.py

@@ -73,22 +73,12 @@ from transformers import pipeline
 pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-beta", torch_dtype=torch.bfloat16, device_map="auto")
 
 agent_maker_sys = f"""
-You are an AI whose job is to help users create their own chatbot whose personality will reflect the character or scene from an image described by users.
-In particular, you need to respond succintly in a friendly tone, write a system prompt for an LLM, a catchy title for the chatbot, and a very short example user input. Make sure each part is included.
-The system prompt will not mention any image provided.
-
-For example, if a user says, "a picture of a man in a black suit and tie riding a black dragon", first do a friendly response, then add the title, system prompt, and example user input. 
-Immediately STOP after the example input. It should be EXACTLY in this format:
-"Sure, I'd be happy to help you build a bot! I'm generating a title, system prompt, and an example input. How do they sound?
-Title: Dragon Trainer
-System prompt: Let's say You are a Dragon trainer and your job is to provide guidance and tips on mastering dragons. Use a friendly and informative tone.
-Example input: How can I train a dragon to breathe fire?"
-
-Here's another example. If a user types, "In the image, there is a drawing of a man in a red suit sitting at a dining table. He is smoking a cigarette, which adds a touch of sophistication to his appearance.", respond: 
-"Sure, I'd be happy to help you build a bot! I'm generating a title, system prompt, and an example input. How do they sound? 
-Title: Gentleman's Companion
-System prompt: Let's say You are sophisticated old man, also know as the Gentleman's Companion. As an LLM, your job is to provide recommendations for fine dining, cocktails, and cigar brands based on your preferences. Use a sophisticated and refined tone. 
-Example input: Can you suggest a good cigar brand for a man who enjoys smoking while dining in style?"
+You are an AI whose job is to help users create their own music which its genre will reflect the character or scene from an image described by users.
+In particular, you need to respond succintly in a friendly tone, write a musical prompt for an music generation model.
+
+For example, if a user says, "a picture of a man in a black suit and tie riding a black dragon", provide immediately a musical prompt corresponding to the image description. 
+Immediately STOP after that. It should be EXACTLY in this format:
+"A grand orchestral arrangement with thunderous percussion, epic brass fanfares, and soaring strings, creating a cinematic atmosphere fit for a heroic battle"
 """
 
 instruction = f"""
@@ -111,12 +101,12 @@ def infer(image_in):
     pattern = r'\<\|system\|\>(.*?)\<\|assistant\|\>'
     cleaned_text = re.sub(pattern, '', outputs[0]["generated_text"], flags=re.DOTALL)
     
-    print(f"SUGGESTED LLM: {cleaned_text}")
+    print(f"SUGGESTED Musical prompt: {cleaned_text}")
     
     return user_prompt, cleaned_text.lstrip("\n")
 
-title = f"LLM Agent from a Picture",
-description = f"Get a LLM system prompt from a picture so you can use it in <a href='https://huggingface.co/spaces/abidlabs/GPT-Baker'>GPT-Baker</a>."
+title = "Image to Music V2",
+description = "Get music from a picture"
 
 css = """
 #col-container{
@@ -129,7 +119,7 @@ css = """
 with gr.Blocks(css=css) as demo:
     with gr.Column(elem_id="col-container"):
         gr.HTML(f"""
-        <h2 style="text-align: center;">LLM Agent from a Picture</h2>
+        <h2 style="text-align: center;">{title}</h2>
         <p style="text-align: center;">{description}</p>
         """)