#!/usr/bin/env python3 """Meta 캐러셀 광고 A-1 (Hook 슬라이드) 하이브리드 이미지 생성. Gemini API로 배경(다이어리+볼펜 오브젝트) 생성 → Playwright HTML 오버레이로 한글 카피 합성. 출력: 1080x1080 PNG """ from __future__ import annotations import base64 import sys import time from pathlib import Path import requests from playwright.sync_api import sync_playwright sys.path.insert(0, str(Path(__file__).parent)) import gcloud_auth # noqa: E402 # ───────────────────────────────────────────────────────────────────────────── # 경로 설정 # ───────────────────────────────────────────────────────────────────────────── OUTPUT_DIR = Path("/home/jay/workspace/output/meta-ads/a-group-v6/pilot") BG_JPEG = OUTPUT_DIR / "_bg_a1_hook.jpg" OUTPUT_PNG = OUTPUT_DIR / "pilot-A1-hook.png" HTML_TEMP = OUTPUT_DIR / "_a1_hook_template.html" FONT_DIR = Path.home() / ".local/share/fonts/Pretendard" # ───────────────────────────────────────────────────────────────────────────── # Gemini 배경 생성 프롬프트 # ───────────────────────────────────────────────────────────────────────────── BG_PROMPT = ( "Cinematic still photograph for a Meta ad. " "Scene: A densely written paper planner lying open on a dark desk surface, " "viewed from slightly above at a low 20-degree angle. " "The planner pages are packed with handwritten Korean-style schedule entries, " "strikethroughs, question marks, and circled items — conveying exhausting daily grind. " "A lone ballpoint pen rests diagonally across the planner. " "Faint ink smudges and indentations visible on the paper surface. " "NO people, NO hands, NO faces — objects only. " "Late-evening atmosphere: fluorescent overhead lights already off, " "only faint ambient blue-grey light seeping from an unseen monitor. " "Color palette: deep charcoal, cold blue-grey, muted navy. " "Zero warmth — no yellows, no oranges in the light. " "The bottom 35% of the frame is darkest — fading into near-black. " "Shallow depth of field, pen in focus, planner edges softly blurred. " "Shot on Sony A7RV, 50mm f/1.8, ISO 3200 — slight film grain visible. " "Mood: silent exhaustion, invisible labor, numbness. " "1:1 square composition 1080x1080. No text, no watermark." ) # ───────────────────────────────────────────────────────────────────────────── # Step 1: Gemini 배경 이미지 생성 # ───────────────────────────────────────────────────────────────────────────── MODEL_ID = "gemini-3.1-flash-image-preview" FALLBACK_MODEL_ID = "gemini-3-pro-image-preview" GEMINI_API_BASE = "https://generativelanguage.googleapis.com/v1beta" GEMINI_SCOPE = "https://www.googleapis.com/auth/generative-language" def generate_background() -> Path: """Gemini API로 배경 이미지를 생성하고 JPEG로 저장합니다.""" if BG_JPEG.exists(): print(f"[배경] 캐시된 배경 이미지 사용: {BG_JPEG}") return BG_JPEG print("[배경] Gemini API 토큰 획득 중...") token = gcloud_auth.get_access_token() print(f"[배경] 토큰 획득 완료 ({len(token)} chars)") url = f"{GEMINI_API_BASE}/models/{MODEL_ID}:generateContent" headers = { "Authorization": f"Bearer {token}", "Content-Type": "application/json", } payload = { "contents": [{"parts": [{"text": BG_PROMPT}]}], "generationConfig": {"responseModalities": ["IMAGE", "TEXT"]}, } print(f"[배경] 이미지 생성 요청 중... (모델: {MODEL_ID})") start = time.time() resp = requests.post(url, headers=headers, json=payload, timeout=300) # fallback to pro image model if flash preview unavailable if resp.status_code in (403, 404): print(f"[배경] 모델 접근 불가 (HTTP {resp.status_code}). Fallback: {FALLBACK_MODEL_ID}") url = f"{GEMINI_API_BASE}/models/{FALLBACK_MODEL_ID}:generateContent" resp = requests.post(url, headers=headers, json=payload, timeout=300) resp.raise_for_status() elapsed = time.time() - start data = resp.json() candidates = data.get("candidates", []) if not candidates: raise RuntimeError(f"candidates 없음: {str(data)[:300]}") parts = candidates[0].get("content", {}).get("parts", []) image_part = next((p for p in parts if "inlineData" in p), None) if image_part is None: texts = [p.get("text", "") for p in parts if "text" in p] raise RuntimeError(f"이미지 데이터 없음. 텍스트: {texts[:2]}") mime = image_part["inlineData"].get("mimeType", "image/jpeg") image_bytes = base64.b64decode(image_part["inlineData"]["data"]) ext = ".jpg" if "jpeg" in mime else ".png" save_path = BG_JPEG.with_suffix(ext) save_path.write_bytes(image_bytes) print(f"[배경] 완료: {save_path.name} ({len(image_bytes):,} bytes, {elapsed:.1f}초)") return save_path # ───────────────────────────────────────────────────────────────────────────── # Step 2: HTML 오버레이 템플릿 생성 # ───────────────────────────────────────────────────────────────────────────── def build_html(bg_path: str) -> str: """A-1 Hook 슬라이드 HTML 오버레이를 빌드합니다.""" return f"""
열심히 했다.
근데 왜 나만 안 되지?

방문 횟수, 전화 통화, 상담 건수—
숫자는 쌓이는데 통장은 그대로다.

""" # ───────────────────────────────────────────────────────────────────────────── # Step 3: Playwright 합성 # ───────────────────────────────────────────────────────────────────────────── def capture_overlay(bg_path: Path) -> None: """HTML 오버레이를 Playwright로 캡처하여 최종 PNG를 저장합니다.""" html_content = build_html(str(bg_path.resolve())) HTML_TEMP.write_text(html_content, encoding="utf-8") print(f"[오버레이] HTML 템플릿 저장: {HTML_TEMP}") print("[오버레이] Playwright 캡처 시작...") with sync_playwright() as p: browser = p.chromium.launch() try: page = browser.new_page(viewport={"width": 1080, "height": 1080}) page.goto(f"file://{HTML_TEMP.resolve()}", wait_until="networkidle") # 폰트 렌더링 + 이미지 로드 대기 page.wait_for_timeout(2500) OUTPUT_PNG.parent.mkdir(parents=True, exist_ok=True) page.screenshot(path=str(OUTPUT_PNG), type="png") finally: browser.close() size = OUTPUT_PNG.stat().st_size print(f"[오버레이] 완료: {OUTPUT_PNG} ({size:,} bytes)") # ───────────────────────────────────────────────────────────────────────────── # Main # ───────────────────────────────────────────────────────────────────────────── def main() -> None: print("=" * 60) print("Meta 캐러셀 A-1 Hook 이미지 생성 시작") print(f"출력: {OUTPUT_PNG}") print("=" * 60) OUTPUT_DIR.mkdir(parents=True, exist_ok=True) # Step 1: 배경 이미지 생성 bg_path = generate_background() # Step 2 & 3: HTML 오버레이 합성 capture_overlay(bg_path) print("\n" + "=" * 60) print(f"완료: {OUTPUT_PNG}") print("=" * 60) if __name__ == "__main__": main()