Update app.py

app.py

@@ -10,13 +10,13 @@ import torch
 from PIL import Image
 from transformers import AutoProcessor, AutoModelForCausalLM, pipeline
 
-# 설치 과정은 생략 가능하며 필요한 경우에만 실행
+# Skip installation process if not needed
 try:
     subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 except:
     print("Flash attention installation skipped")
 
-# OpenAI 클라이언트 초기화 (API 키는 환경 변수에서 가져옴)
+# Initialize OpenAI client (API key from environment variable)
 openai_api_key = os.getenv("OPENAI_API")
 if openai_api_key:
     client = OpenAI(api_key=openai_api_key)
@@ -24,22 +24,22 @@ else:
     print("Warning: OPENAI_API key not found in environment variables")
     client = None
 
-# 번역 모델 추가
+# Add translation model
 translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ko-en")
 
-# Florence 모델 초기화
+# Initialize Florence model
 device = "cuda" if torch.cuda.is_available() else "cpu"
 florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
 florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
 
-# 한글 프롬프트 번역 함수
+# Korean prompt translation function
 def translate_prompt(prompt):
-    if any("\uAC00" <= char <= "\uD7A3" for char in prompt):  # 한글이 포함된 경우
+    if any("\uAC00" <= char <= "\uD7A3" for char in prompt):  # If Korean is included
         translated = translator(prompt, max_length=512)[0]['translation_text']
         return translated
     return prompt
 
-# Florence 캡션 함수
+# Florence caption function
 def florence_caption(image):
     if not isinstance(image, Image.Image):
         image = Image.fromarray(image)
@@ -61,7 +61,7 @@ def florence_caption(image):
     )
     return parsed_answer["<MORE_DETAILED_CAPTION>"]
 
-# JSON 파일 로드 함수
+# JSON file load function
 def load_json_file(file_name):
     file_path = os.path.join("data", file_name)
     try:
@@ -71,7 +71,7 @@ def load_json_file(file_name):
         print(f"Warning: Could not load {file_name}. Using empty list.")
         return []
 
-# JSON 데이터 로드
+# Load JSON data
 ARTFORM = load_json_file("artform.json")
 PHOTO_TYPE = load_json_file("photo_type.json")
 BODY_TYPES = load_json_file("body_types.json")
@@ -91,7 +91,7 @@ COMPOSITION = load_json_file("composition.json")
 POSE = load_json_file("pose.json")
 BACKGROUND = load_json_file("background.json")
 
-# 프롬프트 생성 클래스
+# Prompt generation class
 class PromptGenerator:
     def __init__(self, seed=None):
         self.rng = random.Random(seed)
@@ -171,7 +171,7 @@ class PromptGenerator:
         components = []
         custom = kwargs.get("custom", "")
         if custom:
-            custom = translate_prompt(custom)  # 번역 적용
+            custom = translate_prompt(custom)  # Apply translation
             components.append(custom)
         is_photographer = kwargs.get("artform", "").lower() == "photography" or (
             kwargs.get("artform", "").lower() == "random"
@@ -350,7 +350,7 @@ class OpenAIGenerationNode:
             if not self.client:
                 return "Error: OpenAI API key not found. Please set OPENAI_API environment variable."
                 
-            # 고정 모델: gpt-4.1-mini
+            # Fixed model: gpt-4.1-mini
             openai_model = "gpt-4.1-mini"
 
             default_happy_prompt = """Create a detailed visually descriptive caption of this description, which will be used as a prompt for a text to image AI system (caption only, no instructions like "create an image").Remove any mention of digital artwork or artwork style. Give detailed visual descriptions of the character(s), including ethnicity, skin tone, expression etc. Imagine using keywords for a still for someone who has aphantasia. Describe the image style, e.g. any photographic or art styles / techniques utilized. Make sure to fully describe all aspects of the cinematography, with abundant technical details and visual descriptions. If there is more than one image, combine the elements and characters from all of the images creatively into a single cohesive composition with a single background, inventing an interaction between the characters. Be creative in combining the characters into a single cohesive scene. Focus on two primary characters (or one) and describe an interesting interaction between them, such as a hug, a kiss, a fight, giving an object, an emotional reaction / interaction. If there is more than one background in the images, pick the most appropriate one. Your output is only the caption itself, no comments or extra formatting. The caption is in a single long paragraph. If you feel the images are inappropriate, invent a new scene / characters inspired by these. Additionally, incorporate a specific movie director's visual style and describe the lighting setup in detail, including the type, color, and placement of light sources to create the desired mood and atmosphere. Always frame the scene, including details about the film grain, color grading, and any artifacts or characteristics specific."""
@@ -383,7 +383,7 @@ You are allowed to make up film and branding names, and do them like 80's, 90's
                 char_limit = compression_chars[compression_level]
                 base_prompt += f" Compress the output to be concise while retaining key visual details. MAX OUTPUT SIZE no more than {char_limit} characters."
 
-            # 올바른 OpenAI API 호출 형식
+            # Correct OpenAI API call format
             response = self.client.chat.completions.create(
                 model=openai_model,
                 messages=[
@@ -401,10 +401,10 @@ You are allowed to make up film and branding names, and do them like 80's, 90's
                 top_p=1
             )
             
-            # 응답 추출
+            # Extract response
             output = response.choices[0].message.content
             
-            # 출력 정리
+            # Clean output
             if ": " in output:
                 output = output.split(": ", 1)[1].strip()
             elif output.lower().startswith("here"):
@@ -412,7 +412,7 @@ You are allowed to make up film and branding names, and do them like 80's, 90's
                 if len(sentences) > 1:
                     output = ". ".join(sentences[1:]).strip()
             
-            # 프롬프트 저장
+            # Save prompt
             self.save_prompt(output)
             
             return output
@@ -421,7 +421,7 @@ You are allowed to make up film and branding names, and do them like 80's, 90's
             print(f"An error occurred: {e}")
             return f"Error occurred while processing the request: {str(e)}"
 
-# 개선된 CSS
+# Enhanced CSS
 css = """
 footer {
     visibility: hidden;
@@ -438,7 +438,8 @@ footer {
     display: flex;
     justify-content: center;
     gap: 15px;
-    margin: 20px 0;
+    margin: 20px auto;
+    max-width: 100%;
 }
 .prompt-generator-container {
     background: #f8f9fa;
@@ -487,16 +488,16 @@ def create_interface():
     openai_node = OpenAIGenerationNode()
     
     with gr.Blocks(theme="soft", css=css) as demo:
-        # 헤더
+        # Header
         with gr.Row(elem_classes="main-title"):
             gr.HTML("""
                 <h1>🎨 Flux Prompt Generator</h1>
                 <p style="text-align: center; color: #7f8c8d; font-size: 1.1em;">
-                    한글 입력 지원 | AI 기반 프롬프트 생성기
+                    Korean Input Support | AI-Based Prompt Generator
                 </p>
             """)
         
-        # 배지
+        # Badges
         with gr.Row(elem_classes="badge-container"):
             gr.HTML("""
                 <a href="https://huggingface.co/spaces/openfree/Best-AI" target="_blank">
@@ -507,102 +508,102 @@ def create_interface():
                 </a>
             """)
         
-        # 메인 컨테이너
+        # Main container
         with gr.Row():
-            # 왼쪽 패널 - 설정
+            # Left panel - Settings
             with gr.Column(scale=2):
-                gr.HTML('<div class="section-header">🎯 기본 설정</div>')
-                seed = gr.Slider(0, 30000, label='시드값', step=1, value=random.randint(0,30000))
-                custom = gr.Textbox(label="✏️ 커스텀 프롬프트 (한글 가능)", placeholder="원하는 설명을 입력하세요...")
-                subject = gr.Textbox(label="🎭 주제 (선택사항)", placeholder="예: 아름다운 여성, 귀여운 고양이 등...")
+                gr.HTML('<div class="section-header">🎯 Basic Settings</div>')
+                seed = gr.Slider(0, 30000, label='Seed Value', step=1, value=random.randint(0,30000))
+                custom = gr.Textbox(label="✏️ Custom Prompt (Korean Available)", placeholder="Enter your description...")
+                subject = gr.Textbox(label="🎭 Subject (Optional)", placeholder="e.g., beautiful woman, cute cat, etc...")
                 
-                # 전역 옵션 설정
-                gr.HTML('<div class="section-header">⚡ 빠른 설정</div>')
+                # Global option settings
+                gr.HTML('<div class="section-header">⚡ Quick Settings</div>')
                 global_option = gr.Radio(
                     ["Disabled", "Random", "No Figure Rand"],
-                    label="모든 옵션 일괄 설정:",
+                    label="Apply All Options at Once:",
                     value="Disabled",
-                    info="한 번에 모든 설정을 변경합니다"
+                    info="Change all settings at once"
                 )
                 
-                # 상세 설정들
-                with gr.Accordion("🎨 아트폼 & 사진 타입", open=False):
-                    artform = gr.Dropdown(["disabled", "random"] + ARTFORM, label="아트폼", value="disabled")
-                    photo_type = gr.Dropdown(["disabled", "random"] + PHOTO_TYPE, label="사진 타입", value="disabled")
+                # Detailed settings
+                with gr.Accordion("🎨 Artform & Photo Type", open=False):
+                    artform = gr.Dropdown(["disabled", "random"] + ARTFORM, label="Artform", value="disabled")
+                    photo_type = gr.Dropdown(["disabled", "random"] + PHOTO_TYPE, label="Photo Type", value="disabled")
             
-                with gr.Accordion("👤 캐릭터 설정", open=False):
-                    body_types = gr.Dropdown(["disabled", "random"] + BODY_TYPES, label="체형", value="disabled")
-                    default_tags = gr.Dropdown(["disabled", "random"] + DEFAULT_TAGS, label="기본 태그", value="disabled")
-                    roles = gr.Dropdown(["disabled", "random"] + ROLES, label="역할", value="disabled")
-                    hairstyles = gr.Dropdown(["disabled", "random"] + HAIRSTYLES, label="헤어스타일", value="disabled")
-                    clothing = gr.Dropdown(["disabled", "random"] + CLOTHING, label="의상", value="disabled")
+                with gr.Accordion("👤 Character Settings", open=False):
+                    body_types = gr.Dropdown(["disabled", "random"] + BODY_TYPES, label="Body Type", value="disabled")
+                    default_tags = gr.Dropdown(["disabled", "random"] + DEFAULT_TAGS, label="Default Tags", value="disabled")
+                    roles = gr.Dropdown(["disabled", "random"] + ROLES, label="Role", value="disabled")
+                    hairstyles = gr.Dropdown(["disabled", "random"] + HAIRSTYLES, label="Hairstyle", value="disabled")
+                    clothing = gr.Dropdown(["disabled", "random"] + CLOTHING, label="Clothing", value="disabled")
             
-                with gr.Accordion("🏞️ 장면 설정", open=False):
-                    place = gr.Dropdown(["disabled", "random"] + PLACE, label="장소", value="disabled")
-                    lighting = gr.Dropdown(["disabled", "random"] + LIGHTING, label="조명", value="disabled")
-                    composition = gr.Dropdown(["disabled", "random"] + COMPOSITION, label="구도", value="disabled")
-                    pose = gr.Dropdown(["disabled", "random"] + POSE, label="포즈", value="disabled")
-                    background = gr.Dropdown(["disabled", "random"] + BACKGROUND, label="배경", value="disabled")
+                with gr.Accordion("🏞️ Scene Settings", open=False):
+                    place = gr.Dropdown(["disabled", "random"] + PLACE, label="Place", value="disabled")
+                    lighting = gr.Dropdown(["disabled", "random"] + LIGHTING, label="Lighting", value="disabled")
+                    composition = gr.Dropdown(["disabled", "random"] + COMPOSITION, label="Composition", value="disabled")
+                    pose = gr.Dropdown(["disabled", "random"] + POSE, label="Pose", value="disabled")
+                    background = gr.Dropdown(["disabled", "random"] + BACKGROUND, label="Background", value="disabled")
             
-                with gr.Accordion("🎭 스타일 & 아티스트", open=False):
-                    additional_details = gr.Dropdown(["disabled", "random"] + ADDITIONAL_DETAILS, label="추가 디테일", value="disabled")
-                    photography_styles = gr.Dropdown(["disabled", "random"] + PHOTOGRAPHY_STYLES, label="사진 스타일", value="disabled")
-                    device = gr.Dropdown(["disabled", "random"] + DEVICE, label="촬영 장비", value="disabled")
-                    photographer = gr.Dropdown(["disabled", "random"] + PHOTOGRAPHER, label="사진작가", value="disabled")
-                    artist = gr.Dropdown(["disabled", "random"] + ARTIST, label="아티스트", value="disabled")
-                    digital_artform = gr.Dropdown(["disabled", "random"] + DIGITAL_ARTFORM, label="디지털 아트폼", value="disabled")
+                with gr.Accordion("🎭 Style & Artist", open=False):
+                    additional_details = gr.Dropdown(["disabled", "random"] + ADDITIONAL_DETAILS, label="Additional Details", value="disabled")
+                    photography_styles = gr.Dropdown(["disabled", "random"] + PHOTOGRAPHY_STYLES, label="Photography Style", value="disabled")
+                    device = gr.Dropdown(["disabled", "random"] + DEVICE, label="Camera/Device", value="disabled")
+                    photographer = gr.Dropdown(["disabled", "random"] + PHOTOGRAPHER, label="Photographer", value="disabled")
+                    artist = gr.Dropdown(["disabled", "random"] + ARTIST, label="Artist", value="disabled")
+                    digital_artform = gr.Dropdown(["disabled", "random"] + DIGITAL_ARTFORM, label="Digital Artform", value="disabled")
                 
-                generate_button = gr.Button("🚀 프롬프트 생성", variant="primary", elem_classes="generate-button")
+                generate_button = gr.Button("🚀 Generate Prompt", variant="primary", elem_classes="generate-button")
 
-            # 중간 패널 - 이미지와 출력
+            # Middle panel - Image and output
             with gr.Column(scale=2):
-                with gr.Accordion("🖼️ 이미지 캡션 생성", open=False):
-                    input_image = gr.Image(label="이미지 업로드 (선택사항)", type="pil")
-                    caption_output = gr.Textbox(label="생성된 캡션", lines=3)
+                with gr.Accordion("🖼️ Image Caption Generation", open=False):
+                    input_image = gr.Image(label="Upload Image (Optional)", type="pil")
+                    caption_output = gr.Textbox(label="Generated Caption", lines=3)
                     with gr.Row():
-                        create_caption_button = gr.Button("📝 캡션 생성", variant="secondary")
-                        add_caption_button = gr.Button("➕ 프롬프트에 추가", variant="secondary")
+                        create_caption_button = gr.Button("📝 Generate Caption", variant="secondary")
+                        add_caption_button = gr.Button("➕ Add to Prompt", variant="secondary")
 
-                gr.HTML('<div class="section-header">📋 생성된 프롬프트</div>')
-                output = gr.Textbox(label="메인 프롬프트", lines=4)
-                with gr.Accordion("고급 출력 옵션", open=False):
+                gr.HTML('<div class="section-header">📋 Generated Prompts</div>')
+                output = gr.Textbox(label="Main Prompt", lines=4)
+                with gr.Accordion("Advanced Output Options", open=False):
                     t5xxl_output = gr.Textbox(label="T5XXL", lines=2)
                     clip_l_output = gr.Textbox(label="CLIP L", lines=2)
                     clip_g_output = gr.Textbox(label="CLIP G", lines=2)
             
-            # 오른쪽 패널 - OpenAI
+            # Right panel - OpenAI
             with gr.Column(scale=2):
-                gr.HTML('<div class="section-header">🤖 OpenAI 프롬프트 향상</div>')
-                gr.HTML("<p style='text-align: center; color: #95a5a6;'>사용 모델: gpt-4.1-mini</p>")
+                gr.HTML('<div class="section-header">🤖 OpenAI Prompt Enhancement</div>')
+                gr.HTML("<p style='text-align: center; color: #95a5a6;'>Model Used: gpt-4.1-mini</p>")
                 
                 with gr.Row():
-                    happy_talk = gr.Checkbox(label="😊 Happy Talk", value=True, info="더 상세한 설명")
-                    compress = gr.Checkbox(label="🗜️ 압축", value=True, info="출력 길이 제한")
+                    happy_talk = gr.Checkbox(label="😊 Happy Talk", value=True, info="More detailed description")
+                    compress = gr.Checkbox(label="🗜️ Compress", value=True, info="Limit output length")
                 
                 compression_level = gr.Radio(
                     ["soft", "medium", "hard"], 
-                    label="압축 강도", 
+                    label="Compression Strength", 
                     value="hard",
                     visible=True
                 )
                 
-                poster = gr.Checkbox(label="🎬 영화 포스터 스타일", value=False)
+                poster = gr.Checkbox(label="🎬 Movie Poster Style", value=False)
                 
                 custom_base_prompt = gr.Textbox(
-                    label="🛠️ 커스텀 베이스 프롬프트", 
+                    label="🛠️ Custom Base Prompt", 
                     lines=5,
-                    placeholder="OpenAI에게 특별한 지시사항을 입력하세요..."
+                    placeholder="Enter special instructions for OpenAI..."
                 )
                 
-                generate_text_button = gr.Button("✨ AI로 프롬프트 향상", variant="primary", elem_classes="openai-button")
+                generate_text_button = gr.Button("✨ Enhance with AI", variant="primary", elem_classes="openai-button")
                 
                 text_output = gr.Textbox(
-                    label="🎯 AI 향상 결과", 
+                    label="🎯 AI Enhanced Result", 
                     lines=10,
                     elem_classes="output-container"
                 )
 
-        # 이벤트 핸들러들
+        # Event handlers
         def create_caption(image):
             if image is not None:
                 return florence_caption(image)
@@ -635,14 +636,14 @@ def create_interface():
             outputs=text_output
         )
 
-        # 압축 체크박스에 따른 압축 강도 표시/숨기기
+        # Show/hide compression strength based on compress checkbox
         compress.change(
             lambda x: gr.update(visible=x),
             inputs=[compress],
             outputs=[compression_level]
         )
 
-        # 전역 옵션 변경 함수
+        # Global option change function
         def update_all_options(choice):
             updates = []
             dropdown_list = [
@@ -657,7 +658,7 @@ def create_interface():
                 updates = [gr.update(value="random") for _ in dropdown_list]
             else:  # No Figure Random
                 updates = []
-                # 캐릭터 관련 설정은 disabled
+                # Character-related settings are disabled
                 character_dropdowns = [photo_type, body_types, default_tags, roles, hairstyles, clothing, pose, additional_details]
                 other_dropdowns = [artform, place, lighting, composition, background, photography_styles, device, photographer, artist, digital_artform]