Update app.py

app.py

@@ -1,211 +1,134 @@
 import gradio as gr
 import spaces
 from PIL import Image
-import numpy as np
 from gradio_client import Client, handle_file
 import logging
-import os
 import tempfile
+import os
 
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
-# Try multiple VTON models
-VTON_MODELS = [
-    ("yisol/IDM-VTON", "IDM-VTON"),
-    ("levihsu/OOTDiffusion", "OOTDiffusion"),
-]
+vton_client = None
 
-active_client = None
-active_model = None
-
-def init_vton_client():
-    """Try to connect to available VTON models"""
-    global active_client, active_model
-    
-    for model_id, model_name in VTON_MODELS:
-        try:
-            logger.info(f"Trying to connect to {model_name}...")
-            active_client = Client(model_id)
-            active_model = model_name
-            logger.info(f"✅ Connected to {model_name}")
-            return True
-        except Exception as e:
-            logger.warning(f"Could not connect to {model_name}: {e}")
-            continue
-    
-    logger.error("❌ Could not connect to any VTON model")
-    return False
+def init_client():
+    global vton_client
+    try:
+        logger.info("Connecting to IDM-VTON...")
+        vton_client = Client("yisol/IDM-VTON")
+        logger.info("✅ Connected!")
+        return True
+    except Exception as e:
+        logger.error(f"Failed: {e}")
+        return False
 
-# Initialize
-init_vton_client()
+model_ready = init_client()
 
 @spaces.GPU(duration=180)
-def virtual_tryon(person_image, garment_image, garment_desc, progress=gr.Progress()):
-    """
-    Virtual try-on using IDM-VTON or OOTDiffusion
-    """
-    global active_client, active_model
+def virtual_tryon(person_image, garment_image, progress=gr.Progress()):
+    global vton_client
     
     try:
-        if person_image is None or garment_image is None:
-            raise gr.Error("❌ Please upload both person and garment images!")
+        if not person_image or not garment_image:
+            raise gr.Error("Please upload both images!")
         
-        if active_client is None:
-            logger.info("Reconnecting to VTON service...")
-            if not init_vton_client():
-                raise gr.Error("❌ Virtual Try-On service unavailable. Try again in a moment.")
+        if not vton_client:
+            if not init_client():
+                raise gr.Error("Service unavailable. Try again.")
         
-        logger.info(f"🎯 Using {active_model} for virtual try-on")
-        progress(0.1, desc="📸 Preparing images...")
+        logger.info("Starting try-on...")
+        progress(0.2, desc="Preparing...")
         
-        # Prepare images
         person_img = person_image.convert('RGB')
         garment_img = garment_image.convert('RGB')
         
-        # Save to temp files
-        with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as f:
-            person_img.save(f.name, 'JPEG', quality=95)
-            person_path = f.name
+        person_tmp = tempfile.NamedTemporaryFile(delete=False, suffix='.jpg')
+        garment_tmp = tempfile.NamedTemporaryFile(delete=False, suffix='.jpg')
         
-        with tempfile.NamedTemporaryFile(delete=False, suffix='.jpg') as f:
-            garment_img.save(f.name, 'JPEG', quality=95)
-            garment_path = f.name
+        person_img.save(person_tmp.name, 'JPEG', quality=95)
+        garment_img.save(garment_tmp.name, 'JPEG', quality=95)
         
-        progress(0.3, desc=f"🚀 Processing with {active_model} (60-90s)...")
+        person_tmp.close()
+        garment_tmp.close()
+        
+        progress(0.5, desc="Processing (60-90s)...")
+        
+        result = vton_client.predict(
+            handle_file(person_tmp.name),
+            handle_file(garment_tmp.name),
+            "a garment",
+            True,
+            True,
+            30,
+            42,
+            api_name="/tryon"
+        )
+        
+        progress(0.9, desc="Getting result...")
+        
+        result_path = result[0] if isinstance(result, (tuple, list)) else result
+        result_image = Image.open(result_path)
         
         try:
-            # Call IDM-VTON with correct parameters
-            result = active_client.predict(
-                handle_file(person_path),  # vton_img
-                handle_file(garment_path),  # garm_img
-                garment_desc or "a garment",  # n_cloth
-                True,  # is_checked
-                True,  # is_checked_crop
-                30,  # denoise_steps
-                42,  # seed
-                api_name="/tryon"
-            )
-            
-            progress(0.9, desc="📥 Getting result...")
-            
-            # Handle result
-            if isinstance(result, (list, tuple)):
-                result_path = result[0]
-            else:
-                result_path = result
-            
-            result_image = Image.open(result_path)
-            
-            # Cleanup
-            try:
-                os.unlink(person_path)
-                os.unlink(garment_path)
-            except:
-                pass
-            
-            progress(1.0, desc="✅ Done!")
-            logger.info("✅ Try-on completed!")
-            
-            return result_image
-            
-        except Exception as e:
-            # Cleanup
-            try:
-                os.unlink(person_path)
-                os.unlink(garment_path)
-            except:
-                pass
-            
-            logger.error(f"API call failed: {e}")
-            raise gr.Error(f"⚠️ Try-on failed: {str(e)}\n\nThe service may be busy. Try again in 1-2 minutes.")
-        
-    except gr.Error:
-        raise
+            os.unlink(person_tmp.name)
+            os.unlink(garment_tmp.name)
+        except:
+            pass
+        
+        progress(1.0, desc="Done!")
+        logger.info("✅ Complete!")
+        
+        return result_image
+        
     except Exception as e:
         logger.error(f"Error: {e}")
-        import traceback
-        traceback.print_exc()
         raise gr.Error(f"Failed: {str(e)}")
 
-css = """
-#container {max-width: 1400px; margin: auto;}
-.info-box {
-    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
-    color: white;
-    border-radius: 12px;
-    padding: 25px;
-    margin: 20px 0;
-}
-footer {display: none !important;}
-"""
-
-with gr.Blocks(css=css, title="Mirro Virtual Try-On") as demo:
-    
-    gr.Markdown("""
-    # 👗 Mirro Virtual Try-On API
-    ### 🆓 FREE - Realistic Virtual Try-On
+# UI with Gradio 5.x syntax
+with gr.Blocks(title="Mirro Virtual Try-On", theme=gr.themes.Soft()) as demo:
     
-    Real AI-powered garment transfer (not overlay) using IDM-VTON.
-    """)
+    gr.Markdown("# 👗 Mirro Virtual Try-On\n### 🆓 Free AI-Powered Virtual Try-On")
     
-    if active_client:
-        gr.HTML(f"""
-        <div class="info-box">
-            <h3>✅ Connected to {active_model}</h3>
-            <p>🎨 Production-quality realistic results</p>
-            <p>⏱️ Processing: 60-90 seconds per try-on</p>
-            <p>💰 Cost: FREE (shared resources)</p>
-        </div>
-        """)
+    if model_ready:
+        gr.Markdown("🟢 **Ready!** Upload images to start.")
     else:
-        gr.Markdown("⚠️ **Connecting...** Try generating in a moment.")
+        gr.Markdown("🟡 **Connecting...** Try in a moment.")
     
     with gr.Row():
         with gr.Column():
             gr.Markdown("### 📸 Upload Images")
-            
             person_input = gr.Image(
-                label="👤 Person Photo",
-                type="pil",
-                sources=["upload", "webcam", "clipboard"],
-                height=400
+                label="👤 Person Photo", 
+                type="pil", 
+                sources=["upload", "webcam"]
             )
-            
             garment_input = gr.Image(
-                label="👔 Garment Photo",
-                type="pil",
-                sources=["upload", "webcam", "clipboard"],
-                height=400
-            )
-            
-            garment_desc = gr.Textbox(
-                label="Garment Description (Optional)",
-                placeholder="e.g., blue shirt, red dress",
-                value="a garment"
+                label="👔 Garment Photo", 
+                type="pil", 
+                sources=["upload", "webcam"]
             )
-            
-            with gr.Row():
-                clear_btn = gr.ClearButton([person_input, garment_input, garment_desc])
-                submit_btn = gr.Button("✨ Generate Try-On (FREE)", variant="primary", scale=2)
+            btn = gr.Button("✨ Generate Try-On", variant="primary")
         
         with gr.Column():
             gr.Markdown("### 🎯 Result")
-            output_image = gr.Image(label="Virtual Try-On Result", type="pil", height=850)
+            output_image = gr.Image(label="Virtual Try-On Result", type="pil")
     
-    with gr.Accordion("💡 Tips", open=False):
+    with gr.Accordion("💡 Tips for Best Results", open=False):
         gr.Markdown("""
-        ### For Best Results:
-        - Person facing forward, arms slightly away from body
-        - Clear, well-lit photos
-        - Plain backgrounds preferred
-        - Processing takes 60-90 seconds
-        
-        ### What You Get:
-        - ✅ Realistic garment transfer (NOT overlay)
-        - ✅ Natural fabric draping and wrinkles
-        - ✅ Preserves person's face and pose
-        - ✅ Works with selfies or album photos
+        ### Person Photo:
+        - Face forward, arms slightly away from body
+        - Clear, well-lit photo
+        - Plain background preferred
+        
+        ### Garment Photo:
+        - Clear photo of garment (flat lay or on hanger)
+        - Full garment visible
+        - Good lighting
+        
+        ### Processing:
+        - Takes 60-90 seconds
+        - Realistic results (not overlay)
+        - Preserves face and pose
         """)
     
     with gr.Accordion("📱 iOS API Integration", open=False):
@@ -215,41 +138,56 @@ with gr.Blocks(css=css, title="Mirro Virtual Try-On") as demo:
         POST https://gk2291-mirro-app-server.hf.space/api/predict
 ```
         
-        ### Request Format:
+        ### Request Format (JSON):
 ```json
         {
             "data": [
                 "data:image/jpeg;base64,<person_base64>",
-                "data:image/jpeg;base64,<garment_base64>",
-                "a garment"
+                "data:image/jpeg;base64,<garment_base64>"
             ]
         }
 ```
         
+        ### Response Format (JSON):
+```json
+        {
+            "data": ["data:image/png;base64,<result_base64>"]
+        }
+```
+        
         ### Swift Example:
 ```swift
+        let url = URL(string: "https://gk2291-mirro-app-server.hf.space/api/predict")!
+        var request = URLRequest(url: url)
+        request.httpMethod = "POST"
+        request.setValue("application/json", forHTTPHeaderField: "Content-Type")
+        
         let payload: [String: Any] = [
-            "data": [personBase64, garmentBase64, "a garment"]
+            "data": [personBase64, garmentBase64]
         ]
-        // Send to /api/predict endpoint
+        request.httpBody = try? JSONSerialization.data(withJSONObject: payload)
+        
+        URLSession.shared.dataTask(with: request) { data, response, error in
+            // Handle response
+        }.resume()
 ```
         """)
     
-    submit_btn.click(
+    btn.click(
         fn=virtual_tryon,
-        inputs=[person_input, garment_input, garment_desc],
+        inputs=[person_input, garment_input],
         outputs=output_image,
         api_name="predict"
     )
     
     gr.Markdown("""
     ---
-    <div style="text-align: center; padding: 20px; color: #666;">
+    <div style="text-align: center; color: #666; padding: 20px;">
         <p>🆓 100% FREE | ⚡ Powered by IDM-VTON | 🎨 Production Quality</p>
         <p style="font-size: 0.9em;">Perfect for iOS App MVP</p>
     </div>
     """)
 
 if __name__ == "__main__":
-    demo.queue(max_size=30)
+    demo.queue(max_size=20)
     demo.launch(server_name="0.0.0.0", server_port=7860)