Fix bugs

app.py

@@ -8,11 +8,9 @@ import os
 
 def preprocess(image):
     """
-    Preprocess the image for better OCR results.
-    Converts to grayscale, applies Gaussian blur, and thresholding.
+    Preprocess the image for OCR.
     """
-    img = np.array(image)  # Do NOT mirror here
-    
+    img = np.array(image)
     gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
     blur = cv2.GaussianBlur(gray, (5, 5), 0)
     thresh = cv2.threshold(
@@ -22,35 +20,33 @@ def preprocess(image):
     return thresh
 
 def extract_and_speak(image):
-    """
-    Extract text from the image using pytesseract and generate speech using gTTS.
-    """
     processed = preprocess(image)
     text = pytesseract.image_to_string(processed, lang="eng")
-    
     if text.strip() == "":
         return "No readable text found.", None
-    
     tts = gTTS(text)
     tts.save("output.mp3")
-    
     return text, "output.mp3"
 
-# Gradio interface
-interface = gr.Interface(
-    fn=extract_and_speak,
-    inputs=gr.Image(
-        type="pil",
-        sources=["webcam"],   # Only webcam
-        mirror_webcam=True    # Mirror live webcam feed only
-    ),
-    outputs=[
-        gr.Textbox(label="Extracted Text"),
-        gr.Audio(label="Text-to-Speech Output")
-    ],
-    title="GabAI - AI Assistive Reading System",
-    description="Use your webcam to capture printed text. The preview is mirrored, but text is read from the original image."
-)
+with gr.Blocks() as demo:
+    gr.Markdown("## GabAI - AI Assistive Reading System")
+    gr.Markdown(
+        "Use your webcam to capture printed text. The preview below is mirrored for convenience, "
+        "but OCR uses the original image."
+    )
+    
+    # Webcam input
+    webcam = gr.Image(type="pil", source="webcam")
+    
+    # Mirror the preview using CSS (client-side only)
+    webcam.style(**{"transform": "scaleX(-1)"})
+    
+    # Outputs
+    text_output = gr.Textbox(label="Extracted Text")
+    audio_output = gr.Audio(label="Text-to-Speech Output")
+    
+    # Button to process
+    submit = gr.Button("Read Text")
+    submit.click(fn=extract_and_speak, inputs=webcam, outputs=[text_output, audio_output])
 
-if __name__ == "__main__":
-    interface.launch()
+demo.launch()