Update app.py

app.py

@@ -10,11 +10,11 @@ from io import BytesIO
 from google import genai
 from google.genai import types
 
-# Load environment variables
+# 환경변수 로드
 from dotenv import load_dotenv
 load_dotenv()
 
-# Logging configuration
+# 로깅 설정
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
 logger = logging.getLogger(__name__)
 
@@ -24,84 +24,45 @@ def save_binary_file(file_name, data):
 
 def preprocess_prompt(prompt, image1, image2, image3):
     """
-    Process the prompt and interpret function commands.
+    입력된 프롬프트가 영어로만 되어 있는지 확인하고,
+    단순히 "Please generate the image."를 덧붙여 반환합니다.
     """
-    # Check for image existence
-    has_img1 = image1 is not None
-    has_img2 = image2 is not None
-    has_img3 = image3 is not None
-    
-    # Replace #1, #2, #3 with description (if image is missing, note it)
-    if "#1" in prompt:
-        prompt = prompt.replace("#1", "first image" if has_img1 else "first image (none)")
-    if "#2" in prompt:
-        prompt = prompt.replace("#2", "second image" if has_img2 else "second image (none)")
-    if "#3" in prompt:
-        prompt = prompt.replace("#3", "third image" if has_img3 else "third image (none)")
-    
-    # Interpret function commands
-    if "1. Change Image" in prompt:
-        desc_match = re.search(r'#1 to "(.*?)"', prompt)
-        if desc_match:
-            description = desc_match.group(1)
-            prompt = f"Please change the first image to {description}. Reinterpret it with a new style and mood while preserving its main content."
-        else:
-            prompt = "Please creatively transform the first image into a more vivid and artistic version."
-    
-    elif "2. Remove Text" in prompt:
-        text_match = re.search(r'remove "(.*?)" from #1', prompt)
-        if text_match:
-            text_to_remove = text_match.group(1)
-            prompt = f"Please find and naturally remove the text '{text_to_remove}' from the first image, filling the area with a harmonious background."
-        else:
-            prompt = "Please naturally remove all text from the first image to create a clean look."
-    
-    elif "3. Replace Face" in prompt:
-        prompt = "Please seamlessly replace the face in the first image with the face from the second image. Follow the expression and features of the second image while keeping the rest of the first image intact."
-    
-    elif "4. Change Outfit" in prompt:
-        if "#3" in prompt or "or #3" in prompt:
-            prompt = "Please seamlessly change the outfit in the first image to the outfit from the second or third image. Follow the style and color of the referenced image while maintaining the body proportions and pose of the first image."
-        else:
-            prompt = "Please seamlessly change the outfit in the first image to the outfit from the second image. Follow the style and color of the second image while maintaining the body proportions and pose of the first image."
-    
-    elif "5. Change Background" in prompt:
-        prompt = "Please seamlessly replace the background of the first image with the background from the second image, keeping the main subject intact and harmoniously merging with the new background."
-    
-    elif "6. Blend Images (with product)" in prompt:
-        if "#3" in prompt or "or #3" in prompt:
-            prompt = "Please seamlessly blend the first, second, and third images into one image, ensuring that all key elements are included, especially the product."
-        else:
-            prompt = "Please seamlessly blend the first and second images into one image, ensuring that all key elements are included, especially the product."
-    
-    elif "7. Apply Style" in prompt:
-        prompt = "Please transform the content of the first image into the style of the second image, preserving its main subject and composition while applying the artistic style, colors, and texture of the second image."
-    
-    elif "change to red" in prompt:
-        prompt = "Please change the first image to a red tone. Adjust the overall colors to red hues while maintaining a natural look."
-    
-    prompt += " Please generate the image."
+    # 프롬프트에 한글이 포함되어 있으면 에러 발생
+    if re.search(r'[ㄱ-ㅎ가-힣]', prompt):
+        raise ValueError("Error: Prompt must be in English only.")
+
+    # 불필요한 기능 관련 처리는 제거하고 단순 명령어 추가
+    prompt = prompt.strip() + " Please generate the image."
     return prompt
 
 def generate_with_images(prompt, images):
     """
-    Generate an image via the API call.
+    공식 문서에 기반한 올바른 API 호출 방식 구현
     """
     try:
+        # API 키 확인
         api_key = os.environ.get("GEMINI_API_KEY")
         if not api_key:
             return None, "API key is not set. Please check your environment variables."
             
+        # Gemini 클라이언트 초기화
         client = genai.Client(api_key=api_key)
-        logger.info(f"Starting Gemini API request - Prompt: {prompt}")
         
+        logger.info(f"Gemini API 요청 시작 - 프롬프트: {prompt}")
+        
+        # 컨텐츠 준비
         contents = []
+        
+        # 텍스트 프롬프트 추가
         contents.append(prompt)
+        
+        # 이미지 추가 (이미지가 없으면 텍스트만 전송)
         for idx, img in enumerate(images, 1):
             if img is not None:
                 contents.append(img)
-                logger.info(f"Added image #{idx}")
+                logger.info(f"Image #{idx} added.")
         
+        # 생성 설정 - 공식 문서에 따라 responseModalities 설정
         response = client.models.generate_content(
             model="gemini-2.0-flash-exp-image-generation",
             contents=contents,
@@ -114,12 +75,14 @@ def generate_with_images(prompt, images):
             )
         )
         
+        # 임시 파일 생성
         with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmp:
             temp_path = tmp.name
             
             result_text = ""
             image_found = False
             
+            # 응답 처리
             for part in response.candidates[0].content.parts:
                 if hasattr(part, 'text') and part.text:
                     result_text += part.text
@@ -127,144 +90,113 @@ def generate_with_images(prompt, images):
                 elif hasattr(part, 'inline_data') and part.inline_data:
                     save_binary_file(temp_path, part.inline_data.data)
                     image_found = True
-                    logger.info("Successfully extracted image from response")
+                    logger.info("Image extracted successfully from response.")
             
             if not image_found:
                 return None, f"API did not generate an image. Response text: {result_text}"
                     
+            # 결과 이미지 반환
             result_img = Image.open(temp_path)
             if result_img.mode == "RGBA":
                 result_img = result_img.convert("RGB")
                 
-            return result_img, f"Image successfully generated. {result_text}"
+            return result_img, f"Image generated successfully. {result_text}"
             
     except Exception as e:
-        logger.exception("Error occurred during image generation:")
+        logger.exception("Error during image generation:")
         return None, f"Error occurred: {str(e)}"
 
 def process_images_with_prompt(image1, image2, image3, prompt):
     """
-    Process three images and a prompt.
-    Generates an image based on the prompt even if no images are uploaded,
-    and retries up to 3 times if generation fails.
+    3개의 이미지와 프롬프트를 처리하는 함수.
+    이미지 입력이 없어도 프롬프트만으로 API 호출을 진행합니다.
     """
     try:
+        # 이미지 개수 확인 (이미지 없이도 진행)
         images = [image1, image2, image3]
         valid_images = [img for img in images if img is not None]
         
+        # 프롬프트가 없으면 업로드된 이미지 수에 따라 자동 합성 프롬프트 생성 (영어)
         if not prompt or not prompt.strip():
             if len(valid_images) == 0:
-                prompt = "Please creatively generate an image based on this prompt."
-                logger.info("Automatically generated text-only prompt")
+                prompt = "Please generate an image based on the description."
+                logger.info("Auto prompt generated for no image input.")
             elif len(valid_images) == 1:
                 prompt = "Please creatively transform this image into a more vivid and artistic version."
-                logger.info("Automatically generated single image prompt")
+                logger.info("Auto prompt generated for a single image.")
             elif len(valid_images) == 2:
-                prompt = "Please seamlessly blend these two images into one, integrating the elements harmoniously."
-                logger.info("Automatically generated two-image blending prompt")
+                prompt = "Please seamlessly blend these two images, integrating their elements harmoniously into a single image."
+                logger.info("Auto prompt generated for two images.")
             else:
-                prompt = "Please creatively blend these three images into one cohesive scene, including all main elements."
-                logger.info("Automatically generated three-image blending prompt")
+                prompt = "Please creatively composite these three images, incorporating their key elements into a natural and coherent scene."
+                logger.info("Auto prompt generated for three images.")
         else:
             prompt = preprocess_prompt(prompt, image1, image2, image3)
         
-        max_retries = 3
-        for attempt in range(max_retries):
-            result_img, status = generate_with_images(prompt, valid_images)
-            if result_img is not None:
-                return result_img, status
-            else:
-                logger.info(f"Image generation failed, retrying ({attempt+1}/{max_retries})")
-        return None, f"Failed to generate image. Last status: {status}"
+        # 새로운 API 호출 방식 사용
+        return generate_with_images(prompt, valid_images)
             
     except Exception as e:
-        logger.exception("Error occurred during image processing:")
+        logger.exception("Error during image processing:")
         return None, f"Error occurred: {str(e)}"
 
-def update_prompt_from_function(function_choice):
-    """
-    Returns a prompt template based on the selected function.
-    (Custom text input has been removed.)
-    """
-    function_templates = {
-        "1. Change Image": '#1 to "desired description"',
-        "2. Remove Text": 'remove "text to remove" from #1',
-        "3. Replace Face": "replace the face in #1 with the face from #2",
-        "4. Change Outfit": "change the outfit in #1 to that from #2 or #3",
-        "5. Change Background": "change the background of #1 to the background from #2",
-        "6. Blend Images (with product)": "blend #1 with #2 or #3",
-        "7. Apply Style": "apply the style of #2 to #1"
-    }
-    
-    return function_templates.get(function_choice, "")
-
-# Gradio Interface
+# Gradio 인터페이스 (기능 선택, 기능 적용, 커스텀 텍스트 관련 요소 제거)
 with gr.Blocks() as demo:
     gr.HTML(
         """
         <div style="text-align: center; margin-bottom: 1rem;">
             <h1>Simple Image Generator</h1>
-            <p>You can generate an image by entering only a prompt. Alternatively, you can select a function below.</p>
+            <p>Upload an image (or leave empty) and click generate to create an image based on the English prompt.</p>
         </div>
         """
     )
     
     with gr.Row():
         with gr.Column():
+            # 3개의 이미지 입력 (이미지 없이도 실행 가능)
             with gr.Row():
-                image1_input = gr.Image(type="pil", label="#1", image_mode="RGB")
-                image2_input = gr.Image(type="pil", label="#2", image_mode="RGB")
-                image3_input = gr.Image(type="pil", label="#3", image_mode="RGB")
-            
-            function_dropdown = gr.Dropdown(
-                choices=[
-                    "1. Change Image", 
-                    "2. Remove Text", 
-                    "3. Replace Face", 
-                    "4. Change Outfit",
-                    "5. Change Background", 
-                    "6. Blend Images (with product)", 
-                    "7. Apply Style"
-                ],
-                label="Select Function",
-                value=None
-            )
+                image1_input = gr.Image(type="pil", label="Image 1", image_mode="RGB")
+                image2_input = gr.Image(type="pil", label="Image 2", image_mode="RGB")
+                image3_input = gr.Image(type="pil", label="Image 3", image_mode="RGB")
             
+            # 프롬프트 입력 (영어로만 입력)
             prompt_input = gr.Textbox(
                 lines=3,
-                placeholder="Enter a prompt or leave blank for automatic generation.",
-                label="Prompt (optional)"
+                placeholder="Enter the prompt in English.",
+                label="Prompt (Required: English only)"
             )
             
+            # 생성 버튼
             submit_btn = gr.Button("Generate Image", variant="primary")
         
         with gr.Column():
+            # 결과 출력
             output_image = gr.Image(label="Generated Image")
             output_text = gr.Textbox(label="Status Message")
             prompt_display = gr.Textbox(label="Used Prompt", visible=True)
     
-    # Update prompt textbox when function is selected
-    function_dropdown.change(
-        fn=update_prompt_from_function,
-        inputs=[function_dropdown],
-        outputs=[prompt_input]
-    )
-    
+    # 이미지 생성 버튼 클릭 이벤트
     def process_and_show_prompt(image1, image2, image3, prompt):
+        # 이미지 개수 확인
         images = [image1, image2, image3]
         valid_images = [img for img in images if img is not None]
         
         try:
+            # 만약 사용자가 프롬프트를 입력했다면 영어만 포함되었는지 확인
+            if prompt and re.search(r'[ㄱ-ㅎ가-힣]', prompt):
+                return None, "Error: Prompt must be in English only.", prompt
+            
+            # 프롬프트가 없으면 자동 생성 (영어)
             auto_prompt = prompt
             if not prompt or not prompt.strip():
                 if len(valid_images) == 0:
-                    auto_prompt = "Please creatively generate an image based on this prompt."
+                    auto_prompt = "Please generate an image based on the description."
                 elif len(valid_images) == 1:
                     auto_prompt = "Please creatively transform this image into a more vivid and artistic version."
                 elif len(valid_images) == 2:
-                    auto_prompt = "Please seamlessly blend these two images into one, integrating the elements harmoniously."
+                    auto_prompt = "Please seamlessly blend these two images, integrating their elements harmoniously into a single image."
                 else:
-                    auto_prompt = "Please creatively blend these three images into one cohesive scene, including all main elements."
+                    auto_prompt = "Please creatively composite these three images, incorporating their key elements into a natural and coherent scene."
             else:
                 auto_prompt = preprocess_prompt(prompt, image1, image2, image3)
             
@@ -272,7 +204,7 @@ with gr.Blocks() as demo:
             
             return result_img, status, auto_prompt
         except Exception as e:
-            logger.exception("Error occurred during processing:")
+            logger.exception("Error during processing:")
             return None, f"Error occurred: {str(e)}", prompt
     
     submit_btn.click(
@@ -283,14 +215,14 @@ with gr.Blocks() as demo:
 
     gr.Markdown(
         """
-        ### Usage:
+        ### Instructions:
         
-        1. **Text-Only Generation:** Generate an image without uploading any image by simply entering a prompt.
-        2. **Select Function:** Choose a function from the dropdown and the corresponding template will automatically apply to the prompt.
-        3. **Image Upload:** Upload one or more images to apply the function accordingly.
-        4. **Automatic Retry:** If image generation fails, it will automatically retry up to 2 additional times.
+        1. **Auto Generation**: You can leave the image upload empty and the system will generate an image based solely on the prompt.
+        2. **Prompt Requirement**: Enter the prompt in English only.
+        3. **Image Reference**: The app supports up to three image inputs.
         """
     )
 
+# 애플리케이션 실행
 if __name__ == "__main__":
     demo.launch(share=True)