Update app.py

app.py

@@ -2,11 +2,9 @@ import gradio as gr
 import numpy as np
 from PIL import Image
 import io
-import base64
-import requests
 from huggingface_hub import InferenceClient
 
-# Initialize clients
+# Initialize client
 client = InferenceClient()
 
 def segment_and_edit(image, prompt):
@@ -19,20 +17,26 @@ def segment_and_edit(image, prompt):
         image.save(img_byte_arr, format='PNG')
         img_byte_arr = img_byte_arr.getvalue()
         
-        # Use ControlNet Canny for guided editing (faster than segmentation)
-        # This maintains room structure while applying edits
+        # Use a free, public model for image-to-image
         edited_image = client.image_to_image(
             image=img_byte_arr,
-            prompt=f"interior design photo, {prompt}, professional photography, 4k, detailed",
-            negative_prompt="blurry, distorted, low quality, watermark",
-            model="stabilityai/stable-diffusion-2-1",
-            strength=0.75,  # Balance between original and edit
+            prompt=f"interior design photo, {prompt}, professional photography, high quality, detailed, realistic",
+            negative_prompt="blurry, distorted, low quality, watermark, text, cartoon, sketch",
+            model="black-forest-labs/FLUX.1-schnell",  # Fast, public model
         )
         
         return edited_image
         
     except Exception as e:
-        return None, f"Error: {str(e)}"
+        # Fallback to text-to-image if image-to-image fails
+        try:
+            generated = client.text_to_image(
+                prompt=f"interior room design, {prompt}, professional photography, 4k, realistic",
+                model="black-forest-labs/FLUX.1-schnell"
+            )
+            return generated
+        except Exception as e2:
+            return None, f"Error: {str(e2)}"
 
 def process_room(image, edit_prompt):
     """Main processing function"""