Update app.py

app.py

@@ -1,243 +1,129 @@
 import gradio as gr
 import torch
-from diffusers import AutoPipelineForImage2Image
+from diffusers import StableDiffusionInstructPix2PixPipeline
 from PIL import Image
-import numpy as np
-from transformers import pipeline
 import spaces
 
-# Initialize models
-@spaces.GPU
-def load_models():
-    """Load the image editing models"""
-    try:
-        # Using Stable Diffusion XL for image editing
-        pipe = AutoPipelineForImage2Image.from_pretrained(
-            "stabilityai/stable-diffusion-xl-refiner-1.0",
-            torch_dtype=torch.float16,
-            variant="fp16",
-            use_safetensors=True
-        )
-        pipe.to("cuda")
-        pipe.enable_model_cpu_offload()
-        return pipe
-    except Exception as e:
-        print(f"Error loading model: {e}")
-        return None
-
 # Global model variable
 pipe = None
 
-@spaces.GPU
-def edit_image(input_image, edit_prompt, negative_prompt, strength, guidance_scale, steps):
-    """
-    Edit an image based on text prompt
-    
-    Args:
-        input_image: Input PIL Image
-        edit_prompt: What to change/add in the image
-        negative_prompt: What to avoid in the output
-        strength: How much to change (0-1)
-        guidance_scale: How closely to follow prompt
-        steps: Number of inference steps
-    """
+def load_model():
+    """Load the image editing model"""
     global pipe
-    
     if pipe is None:
-        pipe = load_models()
+        print("Loading model...")
+        pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
+            "timbrooks/instruct-pix2pix",
+            torch_dtype=torch.float16,
+            safety_checker=None
+        )
+        pipe.to("cuda")
+        pipe.enable_attention_slicing()
+        print("Model loaded successfully!")
+    return pipe
+
+@spaces.GPU
+def edit_image(input_image, edit_prompt, num_steps, guidance_scale, image_guidance_scale):
+    """Edit an image based on text instructions"""
     
     if input_image is None:
-        return None, "Please upload an image first!"
+        return None, "❌ Please upload an image first!"
     
     if not edit_prompt or edit_prompt.strip() == "":
-        return None, "Please provide an edit description!"
+        return None, "❌ Please provide edit instructions!"
     
     try:
-        # Resize image if too large
-        max_size = 1024
+        print(f"Starting edit with prompt: {edit_prompt}")
+        
+        # Load model
+        model = load_model()
+        
+        # Resize if too large
+        max_size = 512
         if max(input_image.size) > max_size:
             ratio = max_size / max(input_image.size)
             new_size = tuple(int(dim * ratio) for dim in input_image.size)
             input_image = input_image.resize(new_size, Image.Resampling.LANCZOS)
         
-        # Ensure image dimensions are multiples of 8
+        # Ensure dimensions are multiples of 8
         width = (input_image.width // 8) * 8
         height = (input_image.height // 8) * 8
         input_image = input_image.resize((width, height))
         
+        print(f"Processing image: {width}x{height}")
+        
         # Generate edited image
-        result = pipe(
-            prompt=edit_prompt,
+        result = model(
+            edit_prompt,
             image=input_image,
-            negative_prompt=negative_prompt,
-            strength=strength,
+            num_inference_steps=num_steps,
             guidance_scale=guidance_scale,
-            num_inference_steps=steps,
+            image_guidance_scale=image_guidance_scale,
         ).images[0]
         
+        print("Edit completed successfully!")
         return result, "✅ Image edited successfully!"
         
     except Exception as e:
         error_msg = f"❌ Error: {str(e)}"
         print(error_msg)
+        import traceback
+        traceback.print_exc()
         return None, error_msg
 
-# Example prompts for inspiration
-examples = [
-    ["examples/sample1.jpg", "change the text on the sign to say 'OPEN 24/7'", "blurry, low quality", 0.7, 7.5, 30],
-    ["examples/sample2.jpg", "remove the person in the background", "distorted, artifacts", 0.6, 7.5, 30],
-    ["examples/sample3.jpg", "add sunset lighting and warm tones", "dark, cold colors", 0.5, 7.5, 30],
-    ["examples/sample4.jpg", "make it look like a watercolor painting", "realistic, photographic", 0.8, 8.0, 35],
-]
-
-# Custom CSS for better UI
+# Custom CSS
 custom_css = """
-#main_title {
-    text-align: center;
-    background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
-    -webkit-background-clip: text;
-    -webkit-text-fill-color: transparent;
-    font-size: 2.5em;
-    font-weight: bold;
-    margin-bottom: 0.5em;
-}
-
-#subtitle {
-    text-align: center;
-    color: #666;
-    font-size: 1.2em;
-    margin-bottom: 2em;
-}
-
-.gradio-container {
-    max-width: 1200px !important;
-    margin: auto !important;
-}
-
-#edit_btn {
-    background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
-    border: none;
-    font-size: 1.1em;
-    font-weight: bold;
-}
+.gradio-container {max-width: 1200px !important; margin: auto !important;}
+#title {text-align: center; background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
+-webkit-background-clip: text; -webkit-text-fill-color: transparent; font-size: 2.5em; font-weight: bold;}
 """
 
-# Build the Gradio interface
+# Build interface
 with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
     
-    gr.HTML("<h1 id='main_title'>🎨 AI Image Editing Studio</h1>")
-    gr.HTML("<p id='subtitle'>Transform your images with AI-powered editing • Add/Remove Objects • Edit Text • Style Transfer • And More!</p>")
+    gr.HTML("<h1 id='title'>🎨 AI Image Editor</h1>")
+    gr.Markdown("<p style='text-align: center; font-size: 1.2em;'>Edit images with simple text instructions</p>")
     
     with gr.Row():
-        with gr.Column(scale=1):
-            gr.Markdown("### 📤 Upload & Configure")
-            
-            input_image = gr.Image(
-                label="Upload Image",
-                type="pil",
-                height=400
-            )
+        with gr.Column():
+            input_image = gr.Image(label="Upload Image", type="pil", height=400)
             
             edit_prompt = gr.Textbox(
-                label="✏️ What do you want to change?",
-                placeholder="Example: change the text on the sign to 'OPEN', add a sunset in the background, remove the car, make it anime style...",
+                label="✏️ Edit Instructions",
+                placeholder="Examples:\n- make it a cartoon\n- turn the sky to sunset\n- make it black and white\n- add snow on the ground",
                 lines=3
             )
             
-            negative_prompt = gr.Textbox(
-                label="🚫 What to avoid?",
-                placeholder="Example: blurry, low quality, distorted, artifacts...",
-                value="blurry, low quality, distorted, ugly, bad anatomy",
-                lines=2
-            )
-            
             with gr.Accordion("⚙️ Advanced Settings", open=False):
-                strength = gr.Slider(
-                    label="Edit Strength (how much to change)",
-                    minimum=0.1,
-                    maximum=1.0,
-                    value=0.7,
-                    step=0.05,
-                    info="Lower = subtle changes, Higher = dramatic changes"
-                )
-                
-                guidance_scale = gr.Slider(
-                    label="Guidance Scale (prompt adherence)",
-                    minimum=1.0,
-                    maximum=15.0,
-                    value=7.5,
-                    step=0.5,
-                    info="How closely to follow your prompt"
-                )
-                
-                steps = gr.Slider(
-                    label="Inference Steps (quality)",
-                    minimum=10,
-                    maximum=50,
-                    value=30,
-                    step=5,
-                    info="More steps = better quality but slower"
-                )
-            
-            edit_btn = gr.Button("✨ Edit Image", variant="primary", elem_id="edit_btn", size="lg")
+                num_steps = gr.Slider(10, 100, value=20, step=5, label="Steps (quality)")
+                guidance_scale = gr.Slider(1, 10, value=7.5, step=0.5, label="Text Guidance")
+                image_guidance_scale = gr.Slider(1, 2, value=1.5, step=0.1, label="Image Guidance")
             
-            status_text = gr.Textbox(label="Status", interactive=False, visible=False)
+            edit_btn = gr.Button("✨ Edit Image", variant="primary", size="lg")
         
-        with gr.Column(scale=1):
-            gr.Markdown("### ✨ Result")
-            
-            output_image = gr.Image(
-                label="Edited Image",
-                type="pil",
-                height=400
-            )
-            
-            gr.Markdown("""
-            ### 💡 Tips for Best Results:
-            - Be specific in your edit description
-            - For text editing: mention exact words you want
-            - For removing objects: describe what to replace them with
-            - Adjust strength: 0.5-0.7 for subtle edits, 0.7-0.9 for major changes
-            - Use negative prompts to avoid unwanted artifacts
-            """)
+        with gr.Column():
+            output_image = gr.Image(label="Edited Image", type="pil", height=400)
+            status = gr.Textbox(label="Status", interactive=False)
     
-    # Examples section
-    gr.Markdown("### 🎯 Example Edits (Click to load)")
-    gr.Examples(
-        examples=examples,
-        inputs=[input_image, edit_prompt, negative_prompt, strength, guidance_scale, steps],
-        outputs=[output_image, status_text],
-        fn=edit_image,
-        cache_examples=False
-    )
-    
-    # Footer
     gr.Markdown("""
-    ---
-    ### 🚀 Features:
-    - **Text Editing**: Change text on signs, posters, and images
-    - **Object Manipulation**: Add, remove, or modify objects
-    - **Style Transfer**: Convert to anime, painting, sketch styles
-    - **Smart Editing**: AI understands context and maintains quality
-    
-    ### 📝 Popular Use Cases:
-    - Edit storefront signs and banners
-    - Remove unwanted objects from photos
-    - Change product colors and styles
-    - Create artistic variations of photos
-    - Fix and enhance images
-    
-    ⚡ Powered by Stable Diffusion XL | Made with ❤️ for the AI community
+    ### 💡 Example Prompts:
+    - "make it look like a painting"
+    - "turn day into night"
+    - "make it winter"
+    - "add a smile"
+    - "make it black and white"
+    - "turn it into a cartoon"
     """)
     
-    # Connect the edit button
+    gr.Markdown("---\n⚡ Powered by InstructPix2Pix")
+    
+    # Connect button
     edit_btn.click(
         fn=edit_image,
-        inputs=[input_image, edit_prompt, negative_prompt, strength, guidance_scale, steps],
-        outputs=[output_image, status_text]
+        inputs=[input_image, edit_prompt, num_steps, guidance_scale, image_guidance_scale],
+        outputs=[output_image, status]
     )
 
-# Launch the app
 if __name__ == "__main__":
-    demo.queue(max_size=20)
+    demo.queue(max_size=10)
     demo.launch()