diff --git a/app/home/api/routers/v1/home.py b/app/home/api/routers/v1/home.py
index ff3a601..fa6a6de 100644
--- a/app/home/api/routers/v1/home.py
+++ b/app/home/api/routers/v1/home.py
@@ -38,7 +38,7 @@ from app.utils.common import generate_task_id
 from app.utils.logger import get_logger
 from app.utils.nvMapScraper import NvMapScraper, GraphQLException, URLNotFoundException
 from app.utils.nvMapPwScraper import NvMapPwScraper
-from app.utils.prompts.prompts import marketing_prompt, restaurant_marketing_prompt
+from app.utils.prompts.prompts import marketing_prompt
 from app.utils.address_parser import extract_region_from_address
 from app.utils.autotag import autotag_images
 from config import MEDIA_ROOT
@@ -56,7 +56,7 @@ router = APIRouter()
 
 ## 요청 파라미터
 - **query**: 검색어 (필수)
-- **category**: 카테고리 ("accommodation" | "restaurant", 기본값: 전체)
+- **category**: 카테고리
 
 ## 반환 정보
 - **query**: 검색어
@@ -75,7 +75,7 @@ router = APIRouter()
 async def search_accommodation(
     query: str,
 ) -> AccommodationSearchResponse:
-    """장소 자동완성 검색 (숙박/음식점 등)"""
+    """장소 자동완성 검색"""
     results = await naver_search_client.search_accommodation(
         query=query,
         display=10,
@@ -94,39 +94,42 @@ def _extract_region_from_address(road_address: str | None, jibun_address: str |
     return extract_region_from_address(road_address, jibun_address)
 
 
-class _BizTypeOutput(BaseModel):
-    biz_type: Literal["accommodation", "restaurant", "place"]
+class _IndustryOutput(BaseModel):
+    industry: Literal["stay", "restaurant", "cafe", "salon", "clinic", "fitness", "academy", "attraction"]
 
 
-async def _resolve_biz_type(place_type: str | None, category: str) -> str:
-    """크롤링에서 감지된 place_type을 우선 사용하고, 없으면 AI 분류로 판단."""
-    if place_type in ("accommodation", "restaurant"):
-        return place_type
-    if not category:
-        return "place"
+async def _resolve_industry(category: str, customer_name: str = "") -> str:
+    """업체를 통합 프롬프트의 8개 industry enum 중 하나로 AI 분류.
+
+    분류 근거는 카테고리를 우선 사용하고, 카테고리가 없으면 업체명으로 분류한다.
+    근거가 둘 다 없거나 분류 실패 시 빈 문자열 반환(프롬프트가 범용 처리).
+    """
+    if category:
+        basis_label, basis_value = "네이버 지도 카테고리", category
+    elif customer_name:
+        basis_label, basis_value = "업체명", customer_name
+    else:
+        return ""
     try:
         chatgpt = ChatgptService()
         prompt = (
-            f"다음은 네이버 지도의 업종 카테고리 값입니다: '{category}'\n"
-            "이 업종이 숙박업(accommodation), 음식점(restaurant), 그 외(place) 중 어디에 해당하는지 분류하세요."
+            f"{basis_label}: '{basis_value}'\n"
+            "위 정보를 바탕으로 이 업체를 다음 8개 업종 중 가장 적합한 하나로 분류하세요: "
+            "stay(숙박/펜션/호텔), restaurant(음식점), cafe(카페/디저트), "
+            "salon(미용실/네일/뷰티), clinic(병원/의원/치과), fitness(헬스/필라테스/요가), "
+            "academy(학원/교습소), attraction(관광/체험/액티비티)."
         )
         result = await chatgpt._call_pydantic_output_chat_completion(
             prompt=prompt,
-            output_format=_BizTypeOutput,
+            output_format=_IndustryOutput,
             model="gpt-4o-mini",
             img_url=None,
             image_detail_high=False,
         )
-        return result.biz_type
+        return result.industry
     except Exception as e:
-        logger.warning(f"[biz_type] AI 분류 실패, 기본값 place 반환: {e}")
-        return "place"
-
-
-def _select_marketing_prompt(biz_type: str):
-    if biz_type == "restaurant":
-        return restaurant_marketing_prompt
-    return marketing_prompt
+        logger.warning(f"[industry] AI 분류 실패, 빈 값 반환: {e}")
+        return ""
 
 
 @router.post(
@@ -262,19 +265,19 @@ async def _crawling_logic(
         customer_name = scraper.base_info.get("name", "")
         category = scraper.base_info.get("category", "")
         region = _extract_region_from_address(road_address, jibun_address)
-        biz_type = await _resolve_biz_type(scraper.place_type, category)
+        industry = await _resolve_industry(category, customer_name)
 
         processed_info = ProcessedInfo(
             customer_name=customer_name,
             region=region,
             detail_region_info=road_address or jibun_address or "",
-            biz_type=biz_type,
+            industry=industry,
         )
 
         step2_elapsed = (time.perf_counter() - step2_start) * 1000
         logger.info(
             f"[crawling] Step 2 완료 - {customer_name}, {region}, "
-            f"category={category!r}, place_type={scraper.place_type!r}, biz_type={biz_type} ({step2_elapsed:.1f}ms)"
+            f"category={category!r}, place_type={scraper.place_type!r}, industry={industry!r} ({step2_elapsed:.1f}ms)"
         )
 
         # ========== Step 3: ChatGPT 마케팅 분석 ==========
@@ -288,15 +291,15 @@ async def _crawling_logic(
                 "customer_name": customer_name,
                 "region": region,
                 "detail_region_info": road_address or "",
+                "industry": industry,  # 통합 프롬프트 내부 분기용 업종 enum
             }
 
             # Step 3-2: GPT API 호출 → 구조화된 마케팅 분석 결과 반환
-            selected_prompt = _select_marketing_prompt(biz_type)
             marketing_analysis = await chatgpt_service.generate_structured_output(
-                selected_prompt, input_marketing_data
+                marketing_prompt, input_marketing_data
             )
 
-            # Step 3-3: 분석 결과 DB 저장
+            # Step 3-3: 분석 결과 DB 저장 (industry는 Project로 흐르므로 여기엔 미저장)
             marketing_intel = MarketingIntel(
                 place_id=scraper.place_id,
                 intel_result=marketing_analysis.model_dump(),
@@ -352,7 +355,7 @@ async def _crawling_logic(
         "processed_info": processed_info,
         "marketing_analysis": marketing_analysis,
         "m_id": m_id,
-        "biz_type": biz_type if 'biz_type' in locals() else "place",
+        "industry": industry if 'industry' in locals() else "",
     }
 
 
@@ -390,14 +393,19 @@ async def manual_marketing(
         # Step 2: GPT API 호출 → 마케팅 분석 결과 생성
         # place_id 없이 업체명+주소만으로 분석 (크롤링 없이 직접 입력된 경우)
         chatgpt_service = ChatgptService()
+        # 요청에 industry가 있으면 사용, 없으면 업체명 기반 AI 분류
+        industry = request_body.industry or await _resolve_industry(
+            "", request_body.store_name
+        )
+        processed_info.industry = industry
         input_marketing_data = {
             "customer_name": request_body.store_name,
             "region": region,
             "detail_region_info": request_body.address,
+            "industry": industry,  # 통합 프롬프트 내부 분기용 업종 enum
         }
-        selected_prompt = _select_marketing_prompt(request_body.biz_type)
         marketing_analysis = await chatgpt_service.generate_structured_output(
-            selected_prompt, input_marketing_data
+            marketing_prompt, input_marketing_data
         )
 
         # Step 3: 분석 결과 DB 저장 (place_id=None — 네이버 장소와 연결되지 않음)
@@ -431,7 +439,7 @@ async def manual_marketing(
         processed_info=processed_info,
         marketing_analysis=marketing_analysis,
         m_id=m_id,
-        biz_type=request_body.biz_type,
+        industry=industry,
     )
 
 
diff --git a/app/home/models.py b/app/home/models.py
index 0454108..73e216b 100644
--- a/app/home/models.py
+++ b/app/home/models.py
@@ -114,11 +114,11 @@ class Project(Base):
         comment="마케팅 인텔리전스 결과 정보 저장",
     )
 
-    biz_type: Mapped[str] = mapped_column(
+    industry: Mapped[str] = mapped_column(
         String(50),
         nullable=True,
-        default="place",
-        comment="장소 유형 (place: 기본, accommodation: 숙박, restaurant: 음식점)",
+        default="",
+        comment="업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction)",
     )
 
     language: Mapped[str] = mapped_column(
diff --git a/app/home/schemas/home_schema.py b/app/home/schemas/home_schema.py
index 204363a..a3d317e 100644
--- a/app/home/schemas/home_schema.py
+++ b/app/home/schemas/home_schema.py
@@ -78,7 +78,7 @@ class ProcessedInfo(BaseModel):
     customer_name: str = Field(..., description="고객명/가게명 (base_info.name)")
     region: str = Field(..., description="지역명 (roadAddress에서 추출한 시 이름)")
     detail_region_info: str = Field(..., description="상세 지역 정보 (roadAddress)")
-    biz_type: str = Field(default="place", description="업종 분류 (place: 기본, accommodation: 숙박, restaurant: 음식점)")
+    industry: str = Field(default="", description="업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction)")
 
 
 # class MarketingAnalysisDetail(BaseModel):
@@ -243,7 +243,7 @@ class CrawlingResponse(BaseModel):
         None, description="마케팅 분석 결과 . 실패 시 null"
     )
     m_id : int = Field(..., description="마케팅 분석 결과 ID")
-    biz_type: str = Field(default="place", description="업종 분류 (place: 기본, accommodation: 숙박, restaurant: 음식점)")
+    industry: str = Field(default="", description="업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction)")
 
 
 class ManualMarketingRequest(BaseModel):
@@ -260,7 +260,7 @@ class ManualMarketingRequest(BaseModel):
 
     store_name: str = Field(..., description="업체명 / 브랜드명")
     address: str = Field(..., description="도로명 또는 지번 주소")
-    biz_type: str = Field(default="place", description="업종 분류 (place: 기본, accommodation: 숙박, restaurant: 음식점)")
+    industry: str = Field(default="", description="업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 비우면 업체명 기반 AI 분류")
 
 
 class ErrorResponse(BaseModel):
diff --git a/app/lyric/api/routers/v1/lyric.py b/app/lyric/api/routers/v1/lyric.py
index fb1aaf5..c8ef911 100644
--- a/app/lyric/api/routers/v1/lyric.py
+++ b/app/lyric/api/routers/v1/lyric.py
@@ -47,7 +47,7 @@ from app.utils.prompts.chatgpt_prompt import ChatgptService
 from app.utils.logger import get_logger
 from app.utils.pagination import PaginatedResponse, get_paginated
 
-from app.utils.prompts.prompts import lyric_prompt, restaurant_lyric_prompt
+from app.utils.prompts.prompts import lyric_prompt
 import traceback as tb
 import json
 # 로거 설정
@@ -286,9 +286,11 @@ async def generate_lyric(
             "marketing_intelligence_summary" : json.dumps(marketing_intel.intel_result, ensure_ascii = False),
             "language" : request_body.language,
             "timing_rules" : timing_rules["60s"], # 아직은 선택지 하나
+            "industry": request_body.industry,  # 크롤 응답에서 받아 전달된 업종 enum
         }
 
-        selected_lyric_prompt = restaurant_lyric_prompt if request_body.biz_type == "restaurant" else lyric_prompt
+        # 업종 분기는 프롬프트 내부 {industry}로 처리하므로 단일 프롬프트 사용
+        selected_lyric_prompt = lyric_prompt
 
         step1_elapsed = (time.perf_counter() - step1_start) * 1000
         #logger.debug(f"[generate_lyric] Step 1 완료 - 프롬프트 {len(prompt)}자 ({step1_elapsed:.1f}ms)")
@@ -316,7 +318,7 @@ async def generate_lyric(
                 language=request_body.language,
                 user_uuid=current_user.user_uuid,
                 marketing_intelligence=request_body.m_id,
-                biz_type=request_body.biz_type,
+                industry=request_body.industry,
             )
             session.add(project)
             await session.commit()
diff --git a/app/lyric/schemas/lyric.py b/app/lyric/schemas/lyric.py
index ecf8c76..10479bb 100644
--- a/app/lyric/schemas/lyric.py
+++ b/app/lyric/schemas/lyric.py
@@ -76,7 +76,7 @@ class GenerateLyricRequest(BaseModel):
         description="영상 방향 (horizontal: 가로형, vertical: 세로형)",
     ),
     m_id : Optional[int] = Field(None, description="마케팅 인텔리전스 ID 값")
-    biz_type: str = Field(default="place", description="업종 분류 (place: 기본, accommodation: 숙박, restaurant: 음식점)")
+    industry: str = Field(default="", description="업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction)")
     instrumental: bool = Field(default=False, description="BGM 전용 모드 (가사 생성 안 함)")
 
 
diff --git a/app/lyric/worker/lyric_task.py b/app/lyric/worker/lyric_task.py
index fa84303..722e47a 100644
--- a/app/lyric/worker/lyric_task.py
+++ b/app/lyric/worker/lyric_task.py
@@ -195,6 +195,7 @@ async def generate_subtitle_background(
                 customer_name=customer_name,
                 detail_region_info=store_address,
                 language=language,
+                industry=project.industry,
             )
             pitching_output_list = generated_subtitles.pitching_results
 
diff --git a/app/social/services/seo_service.py b/app/social/services/seo_service.py
index 28cf0dc..f4d7f93 100644
--- a/app/social/services/seo_service.py
+++ b/app/social/services/seo_service.py
@@ -94,8 +94,10 @@ class SeoService:
                 ),
                 "language": project.language,
                 "target_keywords": hashtags,
+                "industry": project.industry or "",  # 크롤 시 분류해 Project에 저장한 업종 enum
             }
 
+            # 업종 분기는 프롬프트 내부 {industry}로 처리하므로 단일 프롬프트 사용
             chatgpt = ChatgptService(timeout=180)
             yt_seo_output = await chatgpt.generate_structured_output(yt_upload_prompt, yt_seo_input_data)
 
diff --git a/app/social/uploader/youtube.py b/app/social/uploader/youtube.py
index d2baf20..fc0b982 100644
--- a/app/social/uploader/youtube.py
+++ b/app/social/uploader/youtube.py
@@ -144,7 +144,7 @@ class YouTubeUploader(BaseSocialUploader):
         body = {
             "snippet": {
                 "title": metadata.title,
-                "description": metadata.description or "",
+                "description": self._sanitize_description(metadata.description),
                 "tags": metadata.tags or [],
                 "categoryId": self._get_category_id(metadata),
             },
@@ -380,6 +380,12 @@ class YouTubeUploader(BaseSocialUploader):
                 )
                 return False
 
+    def _sanitize_description(self, description: str | None) -> str:
+        """YouTube API가 거부하는 문자를 제거합니다. (<, > 포함 시 invalidVideoDescription 오류 발생)"""
+        if not description:
+            return ""
+        return description.replace("<", "").replace(">", "")
+
     def _convert_privacy_status(self, privacy_status: PrivacyStatus) -> str:
         """
         PrivacyStatus를 YouTube API 형식으로 변환
diff --git a/app/utils/autotag.py b/app/utils/autotag.py
index 25e0427..7b65909 100644
--- a/app/utils/autotag.py
+++ b/app/utils/autotag.py
@@ -31,7 +31,7 @@ async def autotag_images(image_url_list : list[str]) -> list[dict]: #tag_list
     
     image_result_tasks = [chatgpt.generate_structured_output(image_autotag_prompt, image_input_data, image_input_data['img_url'], False, silent = True) for image_input_data in image_input_data_list] 
     image_result_list: list[BaseModel | BaseException] = await asyncio.gather(*image_result_tasks, return_exceptions=True)
-    MAX_RETRY = 2 # 하드코딩, 어떻게 처리할지는 나중에
+    MAX_RETRY = 2
     for _ in range(MAX_RETRY):
         failed_idx = [i for i, r in enumerate(image_result_list) if isinstance(r, Exception)]
         print("Failed", failed_idx)
diff --git a/app/utils/creatomate.py b/app/utils/creatomate.py
index e7b51df..776e090 100644
--- a/app/utils/creatomate.py
+++ b/app/utils/creatomate.py
@@ -30,6 +30,7 @@ response = await creatomate.make_creatomate_call(template_id, modifications)
 """
 
 import copy
+import random
 import time
 from enum import StrEnum
 from typing import Literal
@@ -470,10 +471,20 @@ class CreatomateService:
             match template_type:
                 case "image":
                     image_score_list = self.calculate_image_slot_score_multi(taged_image_list, template_component_name)
-                    maximum_idx = image_score_list.index(max(image_score_list))
                     if duplicate:
+                        # 이미지 부족(슬롯 수 > 이미지 수): 재사용이 불가피하므로 슬롯에 가장 적합한 이미지(최고점) 선택
+                        maximum_idx = image_score_list.index(max(image_score_list))
                         selected = taged_image_list[maximum_idx]
                     else:
+                        # 이미지 충분(슬롯 수 <= 이미지 수): 상위 3장 중 점수 비례 확률로 선택해 영상 다양화
+                        # 동일 요청을 반복할 때 매번 같은 영상이 생성되는 문제 방지
+                        sorted_indices = sorted(range(len(image_score_list)), key=lambda i: image_score_list[i], reverse=True)
+                        top_indices = sorted_indices[:min(3, len(sorted_indices))]
+                        weights = [image_score_list[i] for i in top_indices]
+                        if sum(weights) > 0:
+                            maximum_idx = random.choices(top_indices, weights=weights, k=1)[0]
+                        else:
+                            maximum_idx = random.choice(top_indices)
                         selected = taged_image_list.pop(maximum_idx)
                     image_name = selected["image_url"]
                     modifications[template_component_name] =image_name
@@ -679,6 +690,8 @@ class CreatomateService:
         """
         url = f"{self.BASE_URL}/v2/renders"
 
+        source["frame_rate"] = 30
+
         last_error: Exception | None = None
 
         for attempt in range(recovery_settings.CREATOMATE_MAX_RETRIES + 1):
diff --git a/app/utils/prompts/prompts.py b/app/utils/prompts/prompts.py
index 356bf7e..9c8f83b 100644
--- a/app/utils/prompts/prompts.py
+++ b/app/utils/prompts/prompts.py
@@ -1,3 +1,5 @@
+import re
+
 import gspread
 from pydantic import BaseModel
 from google.oauth2.service_account import Credentials
@@ -6,6 +8,9 @@ from app.utils.logger import get_logger
 from app.utils.prompts.schemas import *
 from functools import lru_cache
 
+# {known_field} 형태만 치환하기 위한 패턴. JSON 예시 등 리터럴 중괄호는 보존된다.
+_PLACEHOLDER_RE = re.compile(r"\{(\w+)\}")
+
 logger = get_logger("prompt")
 
 _SCOPES = [
@@ -15,15 +20,33 @@ _SCOPES = [
 _sheet_cache: dict[str, tuple[str, str]] = {}
 
 
+def _get_spreadsheet():
+    creds = Credentials.from_service_account_file(
+        prompt_settings.GOOGLE_SERVICE_ACCOUNT_JSON, scopes=_SCOPES
+    )
+    return gspread.authorize(creds).open_by_key(prompt_settings.PROMPT_SPREADSHEET)
+
+
+def _preload_all_sheets(sheet_names: list[str]):
+    ranges = [f"{name}!B2:B3" for name in sheet_names]
+    response = _get_spreadsheet().values_batch_get(ranges)
+    for value_range in response["valueRanges"]:
+        range_name = value_range.get("range", "")
+        name = range_name.split("!")[0].strip("'")
+        if name not in sheet_names:
+            continue
+        values = value_range.get("values", [])
+        if len(values) < 2 or not values[0] or not values[1]:
+            logger.warning(f"Sheet '{name}' has missing data, skipping preload")
+            continue
+        _sheet_cache[name] = (values[1][0], values[0][0])  # (B3=template, B2=model)
+
+
 def _read_sheet_data(sheet_name: str) -> tuple[str, str]:
     if sheet_name not in _sheet_cache:
-        creds = Credentials.from_service_account_file(
-            prompt_settings.GOOGLE_SERVICE_ACCOUNT_JSON, scopes=_SCOPES
-        )
-        gc = gspread.authorize(creds)
-        ws = gc.open_by_key(prompt_settings.PROMPT_SPREADSHEET).worksheet(sheet_name)
-        values = ws.batch_get(["B2:B3"])[0]  # [[B2], [B3]]
-        _sheet_cache[sheet_name] = (values[1][0], values[0][0])  # (B3=template, B2=model)
+        ws = _get_spreadsheet().worksheet(sheet_name)
+        values = ws.batch_get(["B2:B3"])[0]
+        _sheet_cache[sheet_name] = (values[1][0], values[0][0])
     return _sheet_cache[sheet_name]
 
 
@@ -41,43 +64,43 @@ class Prompt():
         self.prompt_output_class = prompt_output_class
         self.prompt_template, self.prompt_model = _read_sheet_data(sheet_name)
 
-    def _reload_prompt(self):
-        _sheet_cache.pop(self.sheet_name, None)
-        self.prompt_template, self.prompt_model = _read_sheet_data(self.sheet_name)
-
     def build_prompt(self, input_data:dict, silent:bool = False) -> str:
         verified_input = self.prompt_input_class(**input_data)
-        build_template = self.prompt_template
-        build_template = build_template.format(**verified_input.model_dump())
+        values = verified_input.model_dump()
+        # 알려진 {필드}만 치환하고, 프롬프트 내 JSON 예시 등 리터럴 중괄호는 그대로 둔다.
+        # (str.format은 모든 {..}를 필드로 해석해 리터럴 중괄호에서 깨지므로 사용하지 않음)
+        build_template = _PLACEHOLDER_RE.sub(
+            lambda m: str(values[m.group(1)]) if m.group(1) in values else m.group(0),
+            self.prompt_template,
+        )
         if not silent:
             logger.debug(f"build_template: {build_template}")
             logger.debug(f"input_data: {input_data}")
         return build_template
 
+# 업종 분기는 각 프롬프트 내부에서 {industry} 변수로 처리하므로 시트는 타입별 1개로 통합.
+_preload_all_sheets([
+    "marketing",
+    "lyric",
+    "subtitle",
+    "yt_upload",
+    "image_tag",
+])
+
+
+# 업종 분기는 프롬프트 내부 {industry} 변수로 처리하므로 타입별 단일 프롬프트만 둔다.
 marketing_prompt = Prompt(
     sheet_name="marketing",
     prompt_input_class=MarketingPromptInput,
     prompt_output_class=MarketingPromptOutput,
 )
 
-restaurant_marketing_prompt = Prompt(
-    sheet_name="m_restaurant",
-    prompt_input_class=MarketingPromptInput,
-    prompt_output_class=MarketingPromptOutput,
-)
-
 lyric_prompt = Prompt(
     sheet_name="lyric",
     prompt_input_class=LyricPromptInput,
     prompt_output_class=LyricPromptOutput,
 )
 
-restaurant_lyric_prompt = Prompt(
-    sheet_name="l_restaurant",
-    prompt_input_class=LyricPromptInput,
-    prompt_output_class=LyricPromptOutput,
-)
-
 yt_upload_prompt = Prompt(
     sheet_name="yt_upload",
     prompt_input_class=YTUploadPromptInput,
@@ -91,18 +114,10 @@ image_autotag_prompt = Prompt(
 )
 
 @lru_cache()
-def create_dynamic_subtitle_prompt(length: int) -> Prompt:
+def create_dynamic_subtitle_prompt(length: int, industry: str = "") -> Prompt:
+    # industry 인자는 캐시 구분/하위 호환용. 시트는 단일 'subtitle'로 통합됨.
     return Prompt(
         sheet_name="subtitle",
         prompt_input_class=SubtitlePromptInput,
         prompt_output_class=SubtitlePromptOutput[length],
     )
-
-
-def reload_all_prompt():
-    marketing_prompt._reload_prompt()
-    restaurant_marketing_prompt._reload_prompt()
-    lyric_prompt._reload_prompt()
-    restaurant_lyric_prompt._reload_prompt()
-    yt_upload_prompt._reload_prompt()
-    image_autotag_prompt._reload_prompt()
\ No newline at end of file
diff --git a/app/utils/prompts/schemas/image.py b/app/utils/prompts/schemas/image.py
index 71b67d3..4c38a7d 100644
--- a/app/utils/prompts/schemas/image.py
+++ b/app/utils/prompts/schemas/image.py
@@ -94,6 +94,7 @@ class NarrativePreference(BaseModel):
 # Input 정의
 class ImageTagPromptInput(BaseModel):
     img_url : str = Field(..., description="이미지 URL")
+    industry : str = Field(default="", description = "업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 통합 프롬프트가 이 값으로 내부 분기")
     space_type: list[str] = Field(list(SpaceType), description="공간적 정보를 가지는 태그 리스트")
     subject: list[str] = Field(list(Subject), description="피사체 정보를 가지는 태그 리스트")
     camera: list[str] = Field(list(Camera), description="카메라 정보를 가지는 태그 리스트")
diff --git a/app/utils/prompts/schemas/lyric.py b/app/utils/prompts/schemas/lyric.py
index a00e1ba..c2cef87 100644
--- a/app/utils/prompts/schemas/lyric.py
+++ b/app/utils/prompts/schemas/lyric.py
@@ -9,6 +9,7 @@ class LyricPromptInput(BaseModel):
     marketing_intelligence_summary : Optional[str] = Field(None, description = "마케팅 분석 정보 보고서")
     language : str= Field(..., description = "가사 언어")
     timing_rules : str = Field(..., description = "시간 제어문")
+    industry : str = Field(default="", description = "업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 통합 프롬프트가 이 값으로 내부 분기")
 
 # Output 정의
 class LyricPromptOutput(BaseModel):
diff --git a/app/utils/prompts/schemas/marketing.py b/app/utils/prompts/schemas/marketing.py
index 23a4d8c..5dbc906 100644
--- a/app/utils/prompts/schemas/marketing.py
+++ b/app/utils/prompts/schemas/marketing.py
@@ -6,6 +6,7 @@ class MarketingPromptInput(BaseModel):
     customer_name : str = Field(..., description = "마케팅 대상 사업체 이름")
     region : str = Field(..., description = "마케팅 대상 지역")
     detail_region_info : str = Field(..., description = "마케팅 대상 지역 상세")
+    industry : str = Field(default="", description = "업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 통합 프롬프트가 이 값으로 내부 분기")
 
 # Output 정의
 class BrandIdentity(BaseModel):
diff --git a/app/utils/prompts/schemas/subtitle.py b/app/utils/prompts/schemas/subtitle.py
index aa991e4..46491f7 100644
--- a/app/utils/prompts/schemas/subtitle.py
+++ b/app/utils/prompts/schemas/subtitle.py
@@ -10,6 +10,7 @@ class SubtitlePromptInput(BaseModel):
     customer_name : str = Field(..., description = "마케팅 대상 사업체 이름")
     detail_region_info : str = Field(..., description = "마케팅 대상 지역 상세")
     language : str = Field(default="Korean", description="자막 출력 언어 (Korean, English, Chinese, Japanese, Thai, Vietnamese)")
+    industry : str = Field(default="", description = "업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 통합 프롬프트가 이 값으로 내부 분기")
 
 # Output 정의
 class PitchingOutput(BaseModel):
diff --git a/app/utils/prompts/schemas/youtube.py b/app/utils/prompts/schemas/youtube.py
index d81b97d..2acae39 100644
--- a/app/utils/prompts/schemas/youtube.py
+++ b/app/utils/prompts/schemas/youtube.py
@@ -8,6 +8,7 @@ class YTUploadPromptInput(BaseModel):
     marketing_intelligence_summary : Optional[str] = Field(None, description = "마케팅 분석 정보 보고서")
     language : str= Field(..., description = "영상 언어")
     target_keywords: List[str] = Field(..., description="태그 키워드 리스트")
+    industry : str = Field(default="", description = "업종 분류 (stay|restaurant|cafe|salon|clinic|fitness|academy|attraction). 통합 프롬프트가 이 값으로 내부 분기")
 
 # Output 정의
 class YTUploadPromptOutput(BaseModel):
diff --git a/app/utils/subtitles.py b/app/utils/subtitles.py
index 5c6f12a..cfb6369 100644
--- a/app/utils/subtitles.py
+++ b/app/utils/subtitles.py
@@ -16,16 +16,17 @@ class SubtitleContentsGenerator():
     def __init__(self):
         self.chatgpt_service = ChatgptService(timeout=60.0)
 
-    async def generate_subtitle_contents(self, marketing_intelligence : dict[str, Any], pitching_label_list : list[Any], customer_name : str, detail_region_info : str, language : str = "Korean") -> SubtitlePromptOutput:
+    async def generate_subtitle_contents(self, marketing_intelligence : dict[str, Any], pitching_label_list : list[Any], customer_name : str, detail_region_info : str, language : str = "Korean", industry: str = "") -> SubtitlePromptOutput:
         start = time.perf_counter()
         logger.info(
             f"[SubtitleContentsGenerator] START - customer: {customer_name}, "
             f"pitching_count: {len(pitching_label_list)}, "
             f"labels: {pitching_label_list}, "
-            f"language: {language}"
+            f"language: {language}, "
+            f"industry: {industry}"
         )
 
-        dynamic_subtitle_prompt = create_dynamic_subtitle_prompt(len(pitching_label_list))
+        dynamic_subtitle_prompt = create_dynamic_subtitle_prompt(len(pitching_label_list), industry)
         pitching_label_string = "\n".join(pitching_label_list)
         marketing_intel_string = json.dumps(marketing_intelligence, ensure_ascii=False)
         input_data = {
@@ -34,6 +35,7 @@ class SubtitleContentsGenerator():
             "customer_name" : customer_name,
             "detail_region_info" : detail_region_info,
             "language" : language,
+            "industry": industry,  # 가사/영상 파이프라인에서 전달된 업종 enum
             }
 
         logger.info(
diff --git a/docs/database-schema/migration_2026_06_17_biz_type.sql b/docs/database-schema/migration_2026_06_17_biz_type.sql
index bdf923b..af1a4ad 100644
--- a/docs/database-schema/migration_2026_06_17_biz_type.sql
+++ b/docs/database-schema/migration_2026_06_17_biz_type.sql
@@ -2,9 +2,10 @@
 -- Migration: project 테이블에 biz_type 컬럼 추가
 -- Date: 2026-06-17
 -- Description: 업종 구분 컬럼명 biz_type
---              (accommodation: 숙박, restaurant: 음식점 등 업종 확장 예정)
 -- ============================================================
 
 ALTER TABLE project
     ADD COLUMN biz_type VARCHAR(50) NULL DEFAULT 'place' 
     COMMENT '업종 분류 (place: 기본, accommodation: 숙박, restaurant: 음식점)' AFTER id;
+
+ALTER TABLE project CHANGE biz_type industry VARCHAR(50) DEFAULT '';