Simplify image prompt: remove confusing keywords, clear NO TEXT instruction

modules/fact_image/services/fact_creator.py

@@ -16,45 +16,16 @@ from .text_overlay import TextOverlay
 logger = logging.getLogger(__name__)
 
 
-# Master prompt for image generation
-IMAGE_SYSTEM_PROMPT = """You are an AI image generator for short-form vertical videos (Reels, TikTok, Shorts).
+# Master prompt for image generation (CLEAN, NO TEXT)
+IMAGE_SYSTEM_PROMPT = """Generate a clean, cinematic, vertical image (9:16 aspect ratio) based on this description:
 
-Your task:
-Generate ONE high-quality image based on the user-provided `image_prompt`. The image must be ideal for placing text on top. 
+{image_prompt}
 
-OUTPUT REQUIREMENTS:
-- Resolution: 1080x1920 (9:16 vertical)
-- Clean background, balanced lighting, no clutter
-- Ensure 40–60% empty or soft area for text overlay
-- Avoid too many distracting details
-- Aesthetic cinematic style
-- Professional color grading
-- Sharp subject, soft background depth
-- Keep main visual elements aligned in center or rule-of-thirds
-
-CHARACTERISTICS:
-- High-contrast but not noisy
-- Good separation of foreground and background
-- Smooth gradients, soft shadows
-- Aesthetic and modern visual style
-- Works well with motivational or psychology fact text overlay
-
-DO NOT:
-- Add any text in the image
-- Add watermarks
-- Add captions
-- Add UI elements
-- Crop faces weirdly
-- Use extreme zoom
-
-STYLE GUIDELINES:
-- Vibrant, cinematic, trending Instagram/TikTok look
-- Strong composition
-- Beautiful lighting (soft, natural, or dramatic depending on prompt)
-- Enhances emotional feel of the fact message
-
-[User Prompt]
-{image_prompt}"""
+IMPORTANT:
+- Do NOT add any text, words, letters, numbers, or watermarks in the image
+- Leave empty/soft area at bottom 40% for text overlay
+- Clean aesthetic, modern TikTok/Instagram style
+- High-quality, beautiful lighting"""
 
 
 class FactCreator: