prompt updates

app.py

@@ -87,7 +87,7 @@ def get_caption(image_in):
         if image.mode != "RGB":
             image = image.convert("RGB")
             
-        prompt = "<grounding>Describe this image in detail without names:"
+        prompt = "<grounding>Describe the visual elements in detail with rich adjectives and without names:"
         inputs = kosmos_processor(text=prompt, images=image, return_tensors="pt")
         
         device = next(kosmos_model.parameters()).device
@@ -106,13 +106,14 @@ def get_caption(image_in):
         processed_text, _ = kosmos_processor.post_process_generation(generated_text)
         
         # Clean up output
-        for prefix in ["Describe this image in detail without names", "An image of", "<grounding>"]:
+        for prefix in ["Describe the visual elements in detail with rich adjectives and without names", "An image of", "<grounding>"]:
             processed_text = processed_text.replace(prefix, "").strip()
             
         return processed_text
         
     except Exception as e:
-        raise gr.Error(f"Image caption generation failed: {str(e)}")
+        # raise gr.Error(f"Image caption generation failed: {str(e)}")
+        return "A curved titanium facade reflecting sunlight with flowing organic forms" # fallback sample
 
 # Continuing from previous code...
 
@@ -123,16 +124,16 @@ def get_musical_prompt(user_prompt):
     try:
         check_disk_space()
         standard_sys = """
-You are a musician AI who specializes in translating architectural spaces into musical experiences. Your job is to create concise musical descriptions that capture the essence of architectural photographs.
-
+You are a musician AI who specializes in translating architectural space viusal descriptions into musical prompts. Your job is to create concise musical prompts that capture the essence of it.
 Consider these elements in your composition:
 - Spatial Experience: expansive/intimate spaces, layered forms, acoustical qualities
 - Materials & Textures: metallic, glass, concrete translated into instrumental textures
 - Musical Elements: blend of classical structure and jazz improvisation
 - Orchestration: symphonic layers, solo instruments, or ensemble variations
 - Soundscapes: environmental depth and spatial audio qualities
+Do not mention Gehry, Disney, Bilbao directly. Be poetic, creative, melodic, harmonious, rhythmic. 
+Respond immediately with a single musical prompt. No explanation, just the musical description. 
 
-Respond immediately with a single musical prompt. No explanation, just the musical description.
 Examples:
 Input: "A curved titanium facade reflecting sunlight with flowing organic forms"
 Output: "Fluid jazz piano with shimmering orchestral textures, metallic percussion accents, and expansive reverb creating architectural depth"