import os import json import time import random from pathlib import Path from google import genai from google.genai import types GEMINI_API_KEY = os.environ["GEMINI_API_KEY"] client = genai.Client(api_key=GEMINI_API_KEY) # ───────────────────────────────────────────────────────────────────────────── # ORDRE CRITIQUE : # Image 0 = grille overview (ancre tous les variants de créatures) # Images 1-7 = chaque créature individuelle (référence détaillée) # Image 8 = Parent (personnage secondaire) # ───────────────────────────────────────────────────────────────────────────── REFS = { # --- APERCU GLOBAL (Image 0) — poids le plus fort pour la cohérence --- "emotion_grid": "emotion_grid.png", # Image 0 -> All 7 creature forms overview # --- CRÉATURES INDIVIDUELLES (Images 1 à 7) --- "happy": "happy_creature.png", # Image 1 -> JOY FORM (yellow bouncy) "fun": "fun_creature.png", # Image 2 -> AMUSEMENT FORM (orange stretchy) "calm": "calm.png", # Image 3 -> CALM FORM (light blue oval) "sad": "sad_creature.png", # Image 4 -> SADNESS FORM (navy gelatin) "fear": "fear_creature.png", # Image 5 -> FEAR FORM (purple trembling) "anger": "anger_creature.png", # Image 6 -> ANGER FORM (red spiky) "confidence": "confidence_creature.png", # Image 7 -> CONFIDENCE FORM (gold star) # --- PARENT (Image 8) --- "mom": "mom.png", # Image 8 -> Le Parent } CHARACTER_MAPPING = ( "STRICT IDENTITY MAPPING — follow exactly: " "Image 0 = OVERVIEW of all 7 emotion creature forms — use for visual consistency reference. " "Image 1 = JOY CREATURE (Character 1, yellow bouncy form). " "Image 2 = AMUSEMENT CREATURE (Character 2, orange stretchy form). " "Image 3 = CALM CREATURE (Character 3, light blue oval form). " "Image 4 = SADNESS CREATURE (Character 4, navy gelatin form). " "Image 5 = FEAR CREATURE (Character 5, purple trembling form). " "Image 6 = ANGER CREATURE (Character 6, red spiky form). " "Image 7 = CONFIDENCE CREATURE (Character 7, gold star form). " "Image 8 = LE PARENT WOMAN (Character 8). " ) LOCK_PHRASES = ( "Maintain strict visual likeness of every emotion creature as shown in their individual reference image. " "Each creature form has a UNIQUE shape, color, and texture — do not mix or confuse them. " "Ensure character structure, colors, and textures remain 100% identical to the reference sheets. " "Do not alter creature proportions or design details. " ) IMAGE_DESCRIPTIONS = ( "Image 0 = SHAPE-SHIFTER OVERVIEW: one magical creature with 7 emotional forms: " "1-yellow round bouncy (Joy), 2-orange tall stretchy (Amusement), 3-light blue oval velvet (Calm), " "4-navy blue drooping gelatin (Sadness), 5-purple trembling oval (Fear), " "6-red spiky teardrop (Anger), 7-gold star radiant (Confidence). " "Image 1 = JOY CREATURE: bright yellow round ball body, small yellow flame antennae on top, " "big round brown eyes, wide happy smile, tiny stubby arms and legs, golden sparkling particles. " "Image 2 = AMUSEMENT CREATURE: bright orange tall elongated flexible rubbery body, " "small orange antennae on head, big round eyes, huge open grin showing teeth, long rubbery arms and legs. " "Image 3 = CALM CREATURE: light sky-blue soft oval matte velvet body, tiny closed peaceful eyes, " "gentle smile, small stubby arms, floats gently, faint soft blue glow halo. " "Image 4 = SADNESS CREATURE: deep navy blue glossy gelatin ghost-like body, long flowing drape lower body, " "sad droopy puppy eyes, reflective wet glossy surface, mist trail beneath. " "Image 5 = FEAR CREATURE: small purple translucent round oval body, large wide white eyes with " "purple irises, two curly antennae on top, tiny clenched fists, shaking vibrational energy lines around it. " "Image 6 = ANGER CREATURE: bright red spiky teardrop-shaped body with pointed bumps all over, " "orange fiery core glow inside, heavy furrowed brow, angry squinting eyes, steam rising from top. " "Image 7 = CONFIDENCE CREATURE: brilliant metallic gold 5-pointed star shape, arms on hips proudly, " "warm brown eyes, wide proud smile, radiant golden star-burst light aura all around. " "Image 8 = LE PARENT WOMAN: short wavy brown hair in loose updo, warm round brown eyes, " "sage green knitted sweater, gold chain necklace with small green pendant, kind and loving expression. " ) PRENOM = "Leo" def load_references(): loaded = {} for name, path in REFS.items(): if Path(path).exists(): loaded[name] = Path(path).read_bytes() print(f"Reference chargee : {path}") else: loaded[name] = None print(f"Pas de reference : {path}") return loaded def call_with_retry(parts, max_retries=5): for attempt in range(max_retries): try: return client.models.generate_content( model="gemini-3-pro-image-preview", contents=parts, config=types.GenerateContentConfig( response_modalities=["IMAGE", "TEXT"] ) ) except Exception as e: wait = (2 ** attempt) + random.uniform(0, 2) print(f"Tentative {attempt + 1}/{max_retries} echouee : {e}") if attempt < max_retries - 1: print(f"Attente {wait:.1f}s avant retry...") time.sleep(wait) else: raise def generate_page(page_data, style_master, refs, output_dir="pages_emotions"): Path(output_dir).mkdir(exist_ok=True) page_num = page_data["page"] page_type = page_data["type"] prompt = page_data["prompt"].replace("{prenom}", PRENOM) output_path = f"{output_dir}/page_{page_num:02d}.png" format_instruction = ( "IMPORTANT: single full square illustration 2362x2362 pixels, no spread, no border, fills entire frame." if page_type == "simple" else "IMPORTANT: double page spread 4724x2362 pixels, 2:1 ratio, safe center fold, no spine line, fills entire frame." ) full_prompt = style_master + " " + format_instruction + " Scene: " + prompt print(f"Page {page_num:02d}...") text = ( CHARACTER_MAPPING + LOCK_PHRASES + IMAGE_DESCRIPTIONS + f"CHILD CHARACTER {PRENOM}: 5 years old, short slightly wavy brown hair, big round brown eyes, " f"rosy cheeks, soft blue pajamas, white socks. Same face and outfit every scene. " + "Now generate: " + full_prompt ) # Image 0 (grille overview) en tête = poids maximal pour la cohérence # puis chaque créature dans l'ordre (1→7), puis parent (8) parts = [types.Part(text=text)] for key in ["emotion_grid", "happy", "fun", "calm", "sad", "fear", "anger", "confidence", "mom"]: if refs.get(key): parts.append(types.Part( inline_data=types.Blob(data=refs[key], mime_type="image/png") )) response = call_with_retry(parts) for part in response.candidates[0].content.parts: if part.inline_data: with open(output_path, "wb") as f: f.write(part.inline_data.data) print(f"OK : {output_path}") return output_path print(f"ERREUR page {page_num} : aucune image dans la reponse") return None def generate_book(prompts_file="prompts_emotions.json", start_page=1, end_page=None): with open(prompts_file, "r", encoding="utf-8") as f: data = json.load(f) style_master = data["style_master"] pages = data["pages"] if end_page: pages = [p for p in pages if start_page <= p["page"] <= end_page] else: pages = [p for p in pages if p["page"] >= start_page] refs = load_references() print(f"Generation : {len(pages)} pages — prenom : {PRENOM}") results = [] for page_data in pages: path = generate_page(page_data, style_master, refs) results.append({ "page": page_data["page"], "path": path, "success": path is not None }) time.sleep(5) success = sum(1 for r in results if r["success"]) print(f"\nTermine : {success}/{len(pages)} pages generees") with open("book_emotions_results.json", "w") as f: json.dump(results, f, indent=2) return results if __name__ == "__main__": import sys prompts_file = sys.argv[1] if len(sys.argv) > 1 else "prompts_emotions.json" start = int(sys.argv[2]) if len(sys.argv) > 2 else 1 end = int(sys.argv[3]) if len(sys.argv) > 3 else None generate_book(prompts_file, start, end)