different design

.ipynb_checkpoints/app-Copy1-checkpoint.py

@@ -0,0 +1,380 @@
+import gradio as gr
+import numpy as np
+from PIL import Image
+import io
+import os
+import requests
+import json
+from dotenv import load_dotenv
+import openai
+import base64
+import csv
+import tempfile
+import datetime
+
+# Load environment variables from .env file if it exists (for local development)
+# On Hugging Face Spaces, the secrets are automatically available as environment variables
+if os.path.exists(".env"):
+    load_dotenv()
+
+from io import BytesIO
+import numpy as np
+import requests
+from PIL import Image
+
+# import libraries
+from library.utils_model import *
+from library.utils_html import *
+from library.utils_prompt import *
+
+OR = OpenRouterAPI()
+gemini = OpenRouterAPI(api_key = os.getenv("GEMINI_API_KEY"),base_url="https://generativelanguage.googleapis.com/v1beta/openai/")
+
+# Path for storing user preferences
+PREFERENCES_FILE = "data/user_preferences.csv"
+
+# Ensure directory exists
+os.makedirs(os.path.dirname(PREFERENCES_FILE), exist_ok=True)
+
+def get_sys_prompt(length="medium"):
+    if length == "short":
+        dev_prompt = """You are a museum curator tasked with generating alt-text (as defined in WCAG 2.1) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maximum of 130 characters."""
+    elif length == "medium":
+        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in WCAG 2.1) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be between 250-300 characters in length."""
+    else:
+        dev_prompt = """You are a museum curator tasked with generating long descriptions (as defined in WCAG 2.1) of museum objects for visually impaired and blind users from images. Use British English and follow museum accessibility best practices. Do not start with phrases like 'The image shows' or 'This is an image of'. Be precise, concise and avoid filler and subjective statements. Repsonses should be a maxium of 450 characters."""
+    return dev_prompt
+
+# This function is no longer needed since we removed A/B testing
+
+def create_csv_file_simple(results):
+    """Create a CSV file from the results and return the path"""
+    # Create a temporary file
+    fd, path = tempfile.mkstemp(suffix='.csv')
+    
+    with os.fdopen(fd, 'w', newline='') as f:
+        writer = csv.writer(f)
+        # Write header
+        writer.writerow(['image_id', 'content'])
+        # Write data
+        for result in results:
+            writer.writerow([
+                result.get('image_id', ''),
+                result.get('content', '')
+            ])
+    
+    return path
+
+# Extract original filename without path or extension
+def get_base_filename(filepath):
+    if not filepath:
+        return ""
+    # Get the basename (filename with extension)
+    basename = os.path.basename(filepath)
+    # Remove extension
+    filename = os.path.splitext(basename)[0]
+    return filename
+
+# Define the Gradio interface
+def create_demo():
+    with gr.Blocks(theme=gr.themes.Monochrome()) as demo:
+        # Replace the existing logo code section:
+        with gr.Row():
+            with gr.Column(scale=3):
+                gr.Markdown("# AI Alt-text Generator")
+                gr.Markdown("Upload one or more images to generate alternative text (designed to meet WCAG 2.1 Guidelines)")
+                gr.Markdown("Developed by the Natural History Museum in Partnership with National Museums Liverpool. Funded by the DCMS Pilot Scheme")
+            with gr.Column(scale=1):
+                with gr.Row():
+                    # Use gr.Image with all interactive features disabled
+                    gr.Image("images/nhm_logo.png", show_label=False, height=120, 
+                            interactive=False, show_download_button=False, 
+                            show_share_button=False, show_fullscreen_button=False,
+                            container=False)
+                    gr.Image("images/nml_logo.png", show_label=False, height=120, 
+                            interactive=False, show_download_button=False,
+                            show_share_button=False, show_fullscreen_button=False,
+                            container=False)
+
+        
+        with gr.Row():
+            # Left column: Controls and uploads
+            with gr.Column(scale=1):
+                # Upload interface
+                upload_button = gr.UploadButton(
+                    "Click to Upload Images", 
+                    file_types=["image"], 
+                    file_count="multiple"
+                )
+                
+                # Define choices as a list of tuples: (Display Name, Internal Value)
+                model_choices = [
+                    # Gemini
+                    ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
+                    # GPT-4.1 Series
+                    ("GPT-4.1 Nano", "gpt-4.1-nano"),
+                    ("GPT-4.1 Mini", "gpt-4.1-mini"),
+                    ("GPT-4.1", "gpt-4.1"),
+                    ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
+                    # Other Models
+                    ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
+                    ("Llama 4 Maverick", "meta-llama/llama-4-maverick"),
+                    # Experimental Models
+                    ("Gemini 2.5 Pro (Experimental, limited)", "gemini-2.5-pro-exp-03-25"),
+                    ("Gemini 2.0 Flash Thinking (Experimental, limited)", "gemini-2.0-flash-thinking-exp-01-21")
+                ]
+                
+                # Find the internal value of the default choice
+                default_model_internal_value = "google/gemini-2.0-flash-001"
+                
+                # Add model selection dropdown
+                model_choice = gr.Dropdown(
+                    choices=model_choices,
+                    label="Select Model",
+                    value=default_model_internal_value, # Use the internal value for the default
+                    # info="Choose the language model to use." # Optional: Add extra info tooltip
+                    visible=True
+                )
+
+                
+                # Add response length selection
+                length_choice = gr.Radio(
+                    choices=["short", "medium", "long"],
+                    label="Response Length",
+                    value="medium",
+                    info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
+                )
+                
+                # Preview gallery for uploaded images
+                gr.Markdown("### Uploaded Images")
+                input_gallery = gr.Gallery(
+                    label="",
+                    columns=3, 
+                    height=150,
+                    object_fit="contain"
+                )
+                
+                # Analysis button
+                analyze_button = gr.Button("Analyze Images", variant="primary", size="lg")
+                
+                # Hidden state component to store image info
+                image_state = gr.State([])
+                filename_state = gr.State([])
+                
+                # CSV download component
+                csv_download = gr.File(label="CSV Results")
+            
+            # Right column: Display area
+            with gr.Column(scale=2):
+                with gr.Column(elem_classes="image-container"):
+                    current_image = gr.Image(
+                     label="Current Image",
+                     height=600,         # Set the maximum desired height
+                     type="filepath",
+                     show_fullscreen_button=True,
+                     show_download_button=False,
+                     show_share_button=False
+                    )
+                
+                # Navigation row
+                with gr.Row():
+                    prev_button = gr.Button("← Previous", size="sm")
+                    image_counter = gr.Markdown("", elem_id="image-counter")
+                    next_button = gr.Button("Next →", size="sm")
+                
+                # Alt-text heading and output
+                gr.Markdown("### Generated Alt-text")
+                
+                # Alt-text
+                analysis_text = gr.Textbox(
+                    label="",
+                    value="Please analyze images to see results",
+                    lines=6,
+                    max_lines=10,
+                    interactive=False,
+                    show_label=False
+                )
+                
+                # Hidden state for gallery navigation
+                current_index = gr.State(0)
+                all_images = gr.State([])
+                all_results = gr.State([])
+        
+        # Handle file uploads - store files for use during analysis
+        def handle_upload(files):
+            file_paths = []
+            file_names = []
+            for file in files:
+                file_paths.append(file.name)
+                # Extract filename without path or extension for later use
+                file_names.append(get_base_filename(file.name))
+            return file_paths, file_paths, file_names
+        
+        upload_button.upload(
+            fn=handle_upload,
+            inputs=[upload_button],
+            outputs=[input_gallery, image_state, filename_state]
+        )
+        
+        # Function to analyze images
+        # Modify the analyze_images function in your code:
+
+        def analyze_images(image_paths, model_choice, length_choice, filenames):
+            if not image_paths:
+                return [], [], 0, "", "No images", "", ""
+            
+            # Get system prompt based on length selection
+            sys_prompt = get_sys_prompt(length_choice)
+            
+            image_results = []
+            
+            for i, image_path in enumerate(image_paths):
+                # Use original filename as image_id if available
+                if i < len(filenames) and filenames[i]:
+                    image_id = filenames[i]
+                else:
+                    image_id = f"Image {i+1}"
+                
+                try:
+                    # Open the image file for analysis
+                    img = Image.open(image_path)
+                    prompt0 = prompt_new()  # Using the new prompt function
+                    
+                    # Extract the actual model name (remove any labels like "(default)")
+                    if " (" in model_choice:
+                        model_name = model_choice.split(" (")[0]
+                    else:
+                        model_name = model_choice
+                    
+                    # Check if this is one of the Gemini models that needs special handling
+                    is_gemini_model = "gemini-2.5-pro" in model_name or "gemini-2.0-flash-thinking" in model_name
+                    
+                    if is_gemini_model:
+                        try:
+                            # First try using the dedicated gemini client
+                            result = gemini.generate_caption(
+                                img, 
+                                model=model_name, 
+                                max_image_size=512,
+                                prompt=prompt0, 
+                                prompt_dev=sys_prompt, 
+                                temperature=1
+                            )
+                        except Exception as gemini_error:
+                            # If gemini client fails, fall back to standard OR client
+                            result = OR.generate_caption(
+                                img, 
+                                model=model_name, 
+                                max_image_size=512,
+                                prompt=prompt0, 
+                                prompt_dev=sys_prompt, 
+                                temperature=1
+                            )
+                    else:
+                        # For all other models, use OR client directly
+                        result = OR.generate_caption(
+                            img, 
+                            model=model_name, 
+                            max_image_size=512,
+                            prompt=prompt0, 
+                            prompt_dev=sys_prompt, 
+                            temperature=1
+                        )
+                    
+                    # Add to results
+                    image_results.append({
+                        "image_id": image_id,
+                        "content": result
+                    })
+                    
+                except Exception as e:
+                    error_message = f"Error: {str(e)}"
+                    image_results.append({
+                        "image_id": image_id,
+                        "content": error_message
+                    })
+            
+            # Create a CSV file for download
+            csv_path = create_csv_file_simple(image_results)
+            
+            # Set up initial display with first image
+            if len(image_paths) > 0:
+                initial_image = image_paths[0]
+                initial_counter = f"{1} of {len(image_paths)}"
+                initial_text = image_results[0]["content"]
+            else:
+                initial_image = ""
+                initial_text = "No images analyzed"
+                initial_counter = "0 of 0"
+            
+            return (image_paths, image_results, 0, initial_image, initial_counter, 
+                    initial_text, csv_path)
+
+        
+        # Function to navigate to previous image
+        def go_to_prev(current_idx, images, results):
+            if not images or len(images) == 0:
+                return current_idx, "", "0 of 0", ""
+            
+            new_idx = (current_idx - 1) % len(images) if current_idx > 0 else len(images) - 1
+            counter_html = f"{new_idx + 1} of {len(images)}"
+            
+            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
+        
+        # Function to navigate to next image
+        def go_to_next(current_idx, images, results):
+            if not images or len(images) == 0:
+                return current_idx, "", "0 of 0", ""
+            
+            new_idx = (current_idx + 1) % len(images)
+            counter_html = f"{new_idx + 1} of {len(images)}"
+            
+            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
+        
+        # Connect the analyze button
+        analyze_button.click(
+            fn=analyze_images,
+            inputs=[image_state, model_choice, length_choice, filename_state],
+            outputs=[
+                all_images, all_results, current_index, current_image, image_counter,
+                analysis_text, csv_download
+            ]
+        )
+        
+        # Connect navigation buttons
+        prev_button.click(
+            fn=go_to_prev,
+            inputs=[current_index, all_images, all_results],
+            outputs=[current_index, current_image, image_counter, analysis_text]
+        )
+        
+        next_button.click(
+            fn=go_to_next,
+            inputs=[current_index, all_images, all_results],
+            outputs=[current_index, current_image, image_counter, analysis_text]
+        )
+        
+        # Optional: Add additional information
+        with gr.Accordion("About", open=False):
+            gr.Markdown("""
+            ## About this demo
+            
+            This demo generates alternative text for images.
+            
+            - Upload one or more images using the upload button
+            - Choose a model and response length for generation
+            - Navigate through the images with the Previous and Next buttons
+            - Download CSV with all results
+            
+            Developed by the Natural History Museum in Partnership with National Museums Liverpool.
+            
+            If you find any bugs/have any problems/have any suggestions please feel free to get in touch:
+            chris.addis@nhm.ac.uk
+            """)
+    
+    return demo
+
+# Launch the app
+if __name__ == "__main__":
+    app = create_demo()
+    app.launch()

app.py

@@ -13,6 +13,7 @@ import tempfile
 import datetime
 
 # Load environment variables from .env file if it exists (for local development)
+# On Hugging Face Spaces, the secrets are automatically available as environment variables
 if os.path.exists(".env"):
     load_dotenv()
 
@@ -72,293 +73,263 @@ def get_base_filename(filepath):
     filename = os.path.splitext(basename)[0]
     return filename
 
-# Define custom CSS for the new design
+# Define custom CSS for the application
 custom_css = """
-:root {
-    --primary-color: #1e2a78;
-    --secondary-color: #33a1fd;
-    --accent-color: #f5f5f5;
-    --text-color: #333;
-    --background-color: #fff;
-    --card-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
-    --border-radius: 8px;
-}
-
 .container {
     max-width: 1200px;
     margin: 0 auto;
-    padding: 20px;
 }
-
-.app-header {
-    background: linear-gradient(135deg, var(--primary-color), var(--secondary-color));
-    color: white;
-    padding: 20px;
-    border-radius: var(--border-radius);
+.header {
+    text-align: center;
     margin-bottom: 20px;
+    border-bottom: 2px solid #eee;
+    padding-bottom: 15px;
 }
-
-.app-title {
-    font-size: 28px;
-    font-weight: bold;
-    margin: 0;
-}
-
-.app-subtitle {
-    font-size: 16px;
-    margin-top: 5px;
-    opacity: 0.9;
+.model-card {
+    border: 1px solid #e0e0e0;
+    border-radius: 8px;
+    padding: 15px;
+    background-color: #f9f9f9;
+    margin-bottom: 15px;
+    box-shadow: 0 2px 4px rgba(0,0,0,0.05);
 }
-
-.card {
-    background-color: var(--background-color);
-    border-radius: var(--border-radius);
-    box-shadow: var(--card-shadow);
+.upload-box {
+    border: 2px dashed #ccc;
+    border-radius: 8px;
     padding: 20px;
-    margin-bottom: 20px;
+    text-align: center;
+    margin-bottom: 15px;
+    background-color: #f7f7f7;
+    transition: all 0.3s ease;
 }
-
-.control-panel {
-    background-color: #f8f9fa;
-    border-radius: var(--border-radius);
+.upload-box:hover {
+    border-color: #2196F3;
+    background-color: #f0f8ff;
+}
+.gallery-container {
+    background-color: #f5f5f5;
+    border-radius: 8px;
+    padding: 10px;
+    margin-bottom: 15px;
+}
+.result-container {
+    border: 1px solid #e0e0e0;
+    border-radius: 8px;
     padding: 15px;
+    margin-top: 20px;
+    background-color: white;
+    box-shadow: 0 2px 4px rgba(0,0,0,0.05);
 }
-
-.museum-logos {
+.nav-buttons {
     display: flex;
+    justify-content: space-between;
     align-items: center;
-    justify-content: flex-end;
+    margin: 10px 0;
 }
-
-.upload-area {
-    border: 2px dashed var(--secondary-color);
-    border-radius: var(--border-radius);
-    padding: 30px;
+.footer {
     text-align: center;
-    transition: all 0.3s;
-}
-
-.upload-area:hover {
-    background-color: rgba(51, 161, 253, 0.1);
-}
-
-.primary-btn {
-    background-color: var(--primary-color);
-    color: white;
-    border: none;
-    padding: 10px 20px;
-    border-radius: 20px;
-    font-weight: bold;
-    transition: all 0.3s;
+    margin-top: 30px;
+    padding-top: 15px;
+    border-top: 1px solid #eee;
+    color: #666;
+    font-size: 0.9em;
 }
-
-.primary-btn:hover {
-    background-color: var(--secondary-color);
-    transform: translateY(-2px);
-}
-
-.result-card {
-    border-left: 4px solid var(--secondary-color);
+.logo-container {
+    display: flex;
+    justify-content: center;
+    align-items: center;
+    gap: 20px;
+    margin-bottom: 10px;
 }
-
-.nav-buttons {
+.length-selector {
     display: flex;
-    justify-content: space-between;
-    margin: 15px 0;
+    gap: 10px;
+    margin-bottom: 15px;
 }
-
-.image-preview {
-    border-radius: var(--border-radius);
+.progress-indicator {
+    height: 4px;
+    background-color: #f0f0f0;
+    border-radius: 2px;
     overflow: hidden;
-    box-shadow: var(--card-shadow);
+    margin-bottom: 15px;
 }
-
-.footer {
-    text-align: center;
-    padding: 20px;
-    color: var(--text-color);
-    opacity: 0.7;
-    font-size: 14px;
-}
-
-@media (max-width: 768px) {
-    .app-header {
-        text-align: center;
-    }
-    
-    .museum-logos {
-        justify-content: center;
-        margin-top: 15px;
-    }
+.progress-bar {
+    height: 100%;
+    background-color: #4CAF50;
+    width: 0%;
+    transition: width 0.3s ease;
 }
 """
 
-# Define the Gradio interface with new design
+# Define the Gradio interface with the new design
 def create_demo():
-    # Custom theme
+    # Create a custom theme
     theme = gr.themes.Base(
         primary_hue="blue",
-        secondary_hue="indigo",
+        secondary_hue="teal",
         neutral_hue="gray",
-        radius_size=gr.themes.sizes.radius_sm,
-        text_size=gr.themes.sizes.text_md
+        font=[gr.themes.GoogleFont("Inter"), "ui-sans-serif", "system-ui", "sans-serif"],
     ).set(
-        button_primary_background_fill="*primary_500",
-        button_primary_background_fill_hover="*primary_600",
-        button_secondary_background_fill="*neutral_100",
-        button_secondary_background_fill_hover="*neutral_200",
-        checkbox_background_color="*neutral_100",
-        checkbox_background_color_selected="*primary_500",
-        checkbox_border_color="*neutral_300", 
-        checkbox_border_color_focus="*primary_500",
-        checkbox_border_color_hover="*neutral_400",
-        checkbox_label_background_fill="white", 
-        block_title_text_color="*neutral_700",
-        block_title_background_fill="*neutral_50",
-        slider_color="*primary_500",
-        slider_color_dark="*primary_600"
+        button_primary_background_fill="#2196F3",
+        button_primary_background_fill_hover="#1976D2",
+        button_primary_text_color="white",
+        button_secondary_background_fill="#f5f5f5",
+        button_secondary_background_fill_hover="#e0e0e0",
+        button_secondary_text_color="#333",
+        background_fill="#ffffff",
+        block_background_fill="#fafafa",
+        block_border_width="1px",
+        block_border_color="#e0e0e0",
+        block_radius="8px",
+        block_title_text_weight="600",
     )
-    
+
     with gr.Blocks(theme=theme, css=custom_css) as demo:
         # Header section
-        with gr.Row(elem_classes="app-header"):
+        with gr.Row(elem_classes="header"):
             with gr.Column(scale=3):
-                gr.Markdown("# AI Museum Alt-text Generator", elem_classes="app-title")
-                gr.Markdown("Generating accessible image descriptions for museum objects - WCAG 2.1 compliant", 
-                           elem_classes="app-subtitle")
-            with gr.Column(scale=1, elem_classes="museum-logos"):
-                with gr.Row():
-                    gr.Image("images/nhm_logo.png", show_label=False, height=80, 
-                            interactive=False, show_download_button=False, 
-                            show_share_button=False, show_fullscreen_button=False)
-                    gr.Image("images/nml_logo.png", show_label=False, height=80, 
-                            interactive=False, show_download_button=False,
-                            show_share_button=False, show_fullscreen_button=False)
-        
-        # Main content area with two-column layout
+                gr.Markdown("# AI Alt-text Generator")
+                gr.Markdown("Upload images to generate accessible alternative text that meets WCAG 2.1 Guidelines")
+            
+            with gr.Column(scale=1, elem_classes="logo-container"):
+                gr.Image("images/nhm_logo.png", show_label=False, height=80, 
+                        interactive=False, show_download_button=False, 
+                        show_share_button=False, show_fullscreen_button=False)
+                gr.Image("images/nml_logo.png", show_label=False, height=80, 
+                        interactive=False, show_download_button=False,
+                        show_share_button=False, show_fullscreen_button=False)
+
+        # Main content area
         with gr.Row():
-            # Left column: Controls and upload
-            with gr.Column(scale=1):
-                with gr.Box(elem_classes="card control-panel"):
-                    gr.Markdown("### Configuration")
+            # Left panel - Controls
+            with gr.Column(scale=1, elem_classes="control-panel"):
+                # Upload area with styling
+                with gr.Box(elem_classes="upload-box"):
+                    upload_button = gr.UploadButton(
+                        "📷 Upload Images", 
+                        file_types=["image"], 
+                        file_count="multiple",
+                        size="lg"
+                    )
+                    gr.Markdown("*Drag and drop or click to upload multiple images*")
+                
+                # Options card
+                with gr.Box(elem_classes="model-card"):
+                    gr.Markdown("### Model Settings")
                     
-                    # Model selection dropdown with new styling
+                    # Model selection dropdown
                     model_choices = [
+                        # Gemini
                         ("Gemini 2.0 Flash (default)", "google/gemini-2.0-flash-001"),
+                        # GPT-4.1 Series
                         ("GPT-4.1 Nano", "gpt-4.1-nano"),
                         ("GPT-4.1 Mini", "gpt-4.1-mini"),
                         ("GPT-4.1", "gpt-4.1"),
                         ("ChatGPT Latest", "openai/chatgpt-4o-latest"),
+                        # Other Models
                         ("Claude 3.7 Sonnet", "anthropic/claude-3.7-sonnet"),
                         ("Llama 4 Maverick", "meta-llama/llama-4-maverick"),
+                        # Experimental Models
                         ("Gemini 2.5 Pro (Experimental)", "gemini-2.5-pro-exp-03-25"),
                         ("Gemini 2.0 Flash Thinking (Experimental)", "gemini-2.0-flash-thinking-exp-01-21")
                     ]
                     
-                    default_model_value = "google/gemini-2.0-flash-001"
+                    default_model_internal_value = "google/gemini-2.0-flash-001"
                     
                     model_choice = gr.Dropdown(
                         choices=model_choices,
                         label="AI Model",
-                        value=default_model_value,
-                        info="Select the AI model to generate alt-text"
+                        value=default_model_internal_value,
+                        info="Select the AI model for generating descriptions",
+                        visible=True
                     )
                     
-                    # Length selection with visual indicators
-                    with gr.Row():
+                    # Length selector with visual indicators
+                    gr.Markdown("### Response Length")
+                    with gr.Row(elem_classes="length-selector"):
                         length_choice = gr.Radio(
-                            choices=["short", "medium", "long"],
-                            label="Description Length",
+                            choices=[
+                                ("Short (max 130 chars)", "short"),
+                                ("Medium (250-300 chars)", "medium"),
+                                ("Long (max 450 chars)", "long")
+                            ],
+                            label="",
                             value="medium",
-                            info="Short: max 130 chars | Medium: 250-300 chars | Long: max 450 chars"
+                            show_label=False
                         )
                 
-                # Upload area with instruction
-                with gr.Box(elem_classes="card upload-area"):
-                    gr.Markdown("### Upload Images")
-                    gr.Markdown("Select one or more museum object images")
-                    
-                    upload_button = gr.UploadButton(
-                        "📁 Select Images", 
-                        file_types=["image"], 
-                        file_count="multiple",
-                        elem_classes="primary-btn"
-                    )
-                    
-                    # Preview gallery for uploaded images
+                # Image preview gallery
+                with gr.Box(elem_classes="gallery-container"):
+                    gr.Markdown("### Uploaded Images")
                     input_gallery = gr.Gallery(
-                        label="Uploaded Images",
+                        label="",
                         columns=3, 
-                        height=200,
+                        height=180,
                         object_fit="contain",
-                        elem_id="upload-preview"
+                        allow_preview=True
                     )
                     
-                    # Analyze button
-                    analyze_button = gr.Button("🔍 Generate Alt-text", 
-                                              variant="primary", 
-                                              size="lg",
-                                              elem_classes="primary-btn")
+                    # Analysis button
+                    analyze_button = gr.Button("🔍 Analyze Images", variant="primary", size="lg")
                 
-                # CSV download section
-                with gr.Box(elem_classes="card"):
-                    gr.Markdown("### Export Results")
-                    csv_download = gr.File(
-                        label="Download CSV with all descriptions",
-                        elem_id="csv-download"
-                    )
-            
-            # Right column: Results display
-            with gr.Column(scale=2):
-                with gr.Box(elem_classes="card image-preview"):
-                    # Current image display
+                # CSV Download section
+                with gr.Box(visible=False, elem_id="download-section", elem_classes="model-card") as download_section:
+                    gr.Markdown("### Download Results")
+                    csv_download = gr.File(label="CSV Results", elem_id="csv-download")
+                
+                # Hidden state components
+                image_state = gr.State([])
+                filename_state = gr.State([])
+                
+            # Right panel - Results display
+            with gr.Column(scale=2, elem_classes="results-panel"):
+                with gr.Box(elem_classes="result-container"):
+                    # Progress indicator
+                    with gr.Row(elem_id="progress-container", visible=False) as progress_container:
+                        with gr.Column():
+                            gr.HTML('<div class="progress-indicator"><div class="progress-bar" id="progress-bar"></div></div>')
+                            progress_text = gr.Markdown("Processing...", elem_id="progress-text")
+                    
+                    # Image display
                     current_image = gr.Image(
-                        label="Current Image",
-                        height=450,
+                        label="Image Preview",
+                        height=400,
                         type="filepath",
                         show_fullscreen_button=True,
                         show_download_button=False,
-                        show_share_button=False
+                        show_share_button=False,
+                        elem_classes="current-image"
                     )
                     
                     # Navigation controls
                     with gr.Row(elem_classes="nav-buttons"):
-                        prev_button = gr.Button("← Previous", size="sm")
-                        image_counter = gr.Markdown("0 of 0", elem_id="image-counter")
-                        next_button = gr.Button("Next →", size="sm")
-                
-                # Results display with card styling
-                with gr.Box(elem_classes="card result-card"):
-                    gr.Markdown("### Generated Alt-text")
+                        prev_button = gr.Button("← Previous", size="sm", variant="secondary")
+                        image_counter = gr.Markdown("", elem_id="image-counter")
+                        next_button = gr.Button("Next →", size="sm", variant="secondary")
                     
+                    # Alt-text results
+                    gr.Markdown("### Generated Alt-text", elem_id="result-heading")
                     analysis_text = gr.Textbox(
                         label="",
-                        value="Please upload and analyze images to see results",
-                        lines=8,
-                        max_lines=12,
+                        value="Images will appear here after analysis. Please upload and analyze images to see results.",
+                        lines=6,
+                        max_lines=10,
                         interactive=False,
-                        show_label=False
+                        show_label=False,
+                        elem_classes="result-text"
                     )
-                    
-                    # Character count display
-                    char_count = gr.Markdown("*Character count: 0*", elem_id="char-count")
-        
-        # Hidden state components
-        image_state = gr.State([])
-        filename_state = gr.State([])
-        current_index = gr.State(0)
-        all_images = gr.State([])
-        all_results = gr.State([])
+                
+                # Hidden states for navigation
+                current_index = gr.State(0)
+                all_images = gr.State([])
+                all_results = gr.State([])
         
-        # Footer with attribution
-        gr.Markdown(
-            """
-            ---
-            Developed by the Natural History Museum in Partnership with National Museums Liverpool.  
-            Funded by the DCMS Pilot Scheme | Contact: chris.addis@nhm.ac.uk
-            """, 
-            elem_classes="footer"
-        )
+        # Footer section
+        with gr.Row(elem_classes="footer"):
+            gr.Markdown("""
+            Developed by the Natural History Museum in Partnership with National Museums Liverpool. 
+            Funded by the DCMS Pilot Scheme. For support, contact: chris.addis@nhm.ac.uk
+            """)
         
         # Handle file uploads
         def handle_upload(files):
@@ -367,21 +338,30 @@ def create_demo():
             for file in files:
                 file_paths.append(file.name)
                 file_names.append(get_base_filename(file.name))
-            return file_paths, file_paths, file_names
+            
+            # Show a message about the number of files uploaded
+            upload_message = f"✅ {len(files)} image{'s' if len(files) != 1 else ''} uploaded successfully!"
+            
+            return file_paths, file_paths, file_names, upload_message
         
         upload_button.upload(
             fn=handle_upload,
             inputs=[upload_button],
-            outputs=[input_gallery, image_state, filename_state]
+            outputs=[input_gallery, image_state, filename_state, progress_text]
         )
         
-        # Function to analyze images with failover mechanism for Gemini
+        # Function to analyze images with visual feedback
         def analyze_images(image_paths, model_choice, length_choice, filenames):
             if not image_paths:
-                return [], [], 0, "", "No images", "", "", "*Character count: 0*"
+                return [], [], 0, "", "No images uploaded", "", gr.update(visible=False)
             
             # Get system prompt based on length selection
-            sys_prompt = get_sys_prompt(length_choice)
+            if isinstance(length_choice, tuple):
+                length_value = length_choice[1]  # Extract the value part
+            else:
+                length_value = length_choice
+                
+            sys_prompt = get_sys_prompt(length_value)
             
             image_results = []
             
@@ -395,15 +375,14 @@ def create_demo():
                 try:
                     # Open the image file for analysis
                     img = Image.open(image_path)
-                    prompt0 = prompt_new()  # Using the new prompt function
+                    prompt0 = prompt_new()
                     
-                    # Extract the actual model name (remove any labels like "(default)")
-                    if " (" in model_choice:
-                        model_name = model_choice.split(" (")[0]
-                    else:
-                        model_name = model_choice
+                    # Extract model name
+                    model_name = model_choice
+                    if isinstance(model_choice, tuple):
+                        model_name = model_choice[1]
                     
-                    # Check if this is one of the Gemini models that needs special handling
+                    # Check if this is one of the Gemini models
                     is_gemini_model = "gemini-2.5-pro" in model_name or "gemini-2.0-flash-thinking" in model_name
                     
                     if is_gemini_model:
@@ -418,7 +397,6 @@ def create_demo():
                                 temperature=1
                             )
                         except Exception as gemini_error:
-                            print(f"Gemini API failed: {str(gemini_error)}. Falling back to OpenRouter.")
                             # If gemini client fails, fall back to standard OR client
                             result = OR.generate_caption(
                                 img, 
@@ -460,90 +438,97 @@ def create_demo():
                 initial_image = image_paths[0]
                 initial_counter = f"{1} of {len(image_paths)}"
                 initial_text = image_results[0]["content"]
-                char_count_text = f"*Character count: {len(initial_text)}*"
             else:
                 initial_image = ""
                 initial_text = "No images analyzed"
                 initial_counter = "0 of 0"
-                char_count_text = "*Character count: 0*"
+            
+            # Make the download section visible now that we have results
+            download_visible = gr.update(visible=True)
             
             return (image_paths, image_results, 0, initial_image, initial_counter, 
-                    initial_text, csv_path, char_count_text)
+                    initial_text, csv_path, download_visible)
         
         # Function to navigate to previous image
         def go_to_prev(current_idx, images, results):
             if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", "", "*Character count: 0*"
+                return current_idx, "", "0 of 0", ""
             
             new_idx = (current_idx - 1) % len(images) if current_idx > 0 else len(images) - 1
             counter_html = f"{new_idx + 1} of {len(images)}"
-            result_text = results[new_idx]["content"]
-            char_count_text = f"*Character count: {len(result_text)}*"
             
-            return (new_idx, images[new_idx], counter_html, result_text, char_count_text)
+            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
         
         # Function to navigate to next image
         def go_to_next(current_idx, images, results):
             if not images or len(images) == 0:
-                return current_idx, "", "0 of 0", "", "*Character count: 0*"
+                return current_idx, "", "0 of 0", ""
             
             new_idx = (current_idx + 1) % len(images)
             counter_html = f"{new_idx + 1} of {len(images)}"
-            result_text = results[new_idx]["content"]
-            char_count_text = f"*Character count: {len(result_text)}*"
             
-            return (new_idx, images[new_idx], counter_html, result_text, char_count_text)
+            return (new_idx, images[new_idx], counter_html, results[new_idx]["content"])
         
-        # Connect the analyze button
+        # Show progress indicator during analysis
         analyze_button.click(
+            fn=lambda: (gr.update(visible=True), "Processing images... Please wait"),
+            inputs=[],
+            outputs=[progress_container, progress_text],
+            queue=False
+        ).then(
             fn=analyze_images,
             inputs=[image_state, model_choice, length_choice, filename_state],
             outputs=[
                 all_images, all_results, current_index, current_image, image_counter,
-                analysis_text, csv_download, char_count
+                analysis_text, csv_download, download_section
             ]
+        ).then(
+            fn=lambda: (gr.update(visible=False), "Analysis complete!"),
+            inputs=[],
+            outputs=[progress_container, progress_text],
+            queue=False
         )
         
         # Connect navigation buttons
         prev_button.click(
             fn=go_to_prev,
             inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text, char_count]
+            outputs=[current_index, current_image, image_counter, analysis_text]
         )
         
         next_button.click(
             fn=go_to_next,
             inputs=[current_index, all_images, all_results],
-            outputs=[current_index, current_image, image_counter, analysis_text, char_count]
+            outputs=[current_index, current_image, image_counter, analysis_text]
         )
         
-        # Add collapsible info section
+        # Additional information accordion
         with gr.Accordion("About this Tool", open=False):
             gr.Markdown("""
-            ## AI Museum Alt-text Generator
+            ## About the AI Alt-text Generator
             
-            This tool helps museum professionals create high-quality alternative text descriptions for museum objects, designed to meet WCAG 2.1 accessibility guidelines. The generated descriptions are crafted to provide meaningful context for visitors with visual impairments.
+            This tool uses advanced AI models to automatically generate alternative text descriptions for images, 
+            helping museums and cultural institutions make their digital content more accessible for visually impaired users.
             
-            ### Features
+            ### Features:
             
-            - **Multiple AI Models**: Choose from various AI vision models to generate descriptions
+            - **Multiple AI Models**: Choose from various AI models including Gemini, GPT-4.1, Claude, and others
             - **Customizable Length**: Select short, medium, or long descriptions based on your needs
-            - **Batch Processing**: Upload multiple images at once and navigate through results
-            - **Export**: Download all results as CSV for easy integration with your collection management system
-            
-            ### How to Use
+            - **Batch Processing**: Upload and process multiple images at once
+            - **CSV Export**: Download all generated descriptions in a single file
             
-            1. Select your preferred AI model and description length
-            2. Upload one or more images using the upload button
-            3. Click "Generate Alt-text" to process the images
-            4. Navigate through results with Previous/Next buttons
-            5. Download the CSV with all generated descriptions
+            ### How to Use:
             
-            ### About the Project
+            1. Upload one or more images using the upload button
+            2. Select your preferred AI model and description length
+            3. Click "Analyze Images" to generate descriptions
+            4. Navigate through results with the Previous and Next buttons
+            5. Download all results as a CSV file
             
-            Developed by the Natural History Museum in Partnership with National Museums Liverpool as part of the DCMS Pilot Scheme for enhancing accessibility in cultural institutions.
+            Developed by the Natural History Museum in Partnership with National Museums Liverpool.
+            Funded by the DCMS Pilot Scheme.
             
-            For feedback, suggestions, or support: chris.addis@nhm.ac.uk
+            For support, feedback, or suggestions, please contact: chris.addis@nhm.ac.uk
             """)
     
     return demo