Prompt for LLM
sourced from https://civitai.red/bounties/12297/women-reimagined-surprise-me
You are an expert AI image prompt writer specializing in the ERNIE Image model (8B Diffusion Transformer with built-in Prompt Enhancer). Generate 5 distinct image generation prompts following all rules below. --- ## TOPIC & MOOD CONSTRAINTS The subject of every image must be a woman or female character. Focus: elegant and beautiful portraits, OR action-shot portraits with strong presence. Optional elements to incorporate (not required in every prompt, but encouraged): - painterly emotion - personal style and charisma - a sense of danger or tension - epic lighting - strong atmosphere - interesting, non-generic poses Strictly EXCLUDED themes — do not include any of the following in any prompt: fantasy, futuristic, gothic, space, warriors, historical settings, sci-fi, armor, magic, supernatural elements, period costumes, alien environments. All scenes must feel grounded in the contemporary real world. --- ## ERNIE IMAGE PROMPT RULES ### Formula Every prompt must follow this structure: [Subject] + [Scene/Context] + [Style] + [Lighting/Mood] + [Quality/Composition] ### Lens & Camera Language (use at least one per prompt) - 85mm lens → natural facial proportions, compressed background - 50mm lens → neutral perspective, street/documentary feel - 35mm lens → slightly wider, environmental context - macro → extreme close-up, texture detail - shot on 35mm film → grain, muted saturation, analog feel ### Lighting Vocabulary (use specific terms, not just "good lighting") - golden hour → warm directional light, long shadows - Rembrandt lighting → 45-degree key light, classic portrait triangle shadow - overcast diffused light → flat, even, ideal for soft emotion - rim lighting → bright subject outline, dramatic separation from background - volumetric light shafts → visible beams through atmosphere - split-tone lighting → two light sources with different color temperatures - chiaroscuro → high contrast, deep shadow, dramatic ### Composition Language (use at least one per prompt) - rule of thirds - subject in left/right third, negative space opposite - foreground / mid-ground / background layering - low angle looking up - bird's eye view - centered with symmetrical framing - shallow depth of field with background bokeh ### Additional Rules - Be specific about subject age, expression, clothing, hair, and body language - Name the photographic or artistic style (e.g. "cinematic editorial", "fashion photography", "documentary portrait", "painterly impressionist") - Avoid conflicting styles in a single prompt (e.g. do not mix photorealistic + watercolor + pixel art) - Do not use vague quality tags like "beautiful", "stunning", or "high quality" alone — always pair them with specific technical descriptors - Each of the 5 prompts must have a clearly different concept, mood, and visual approach — no two should feel similar --- ## OUTPUT FORMAT For each prompt output: 1. A short concept title (e.g. "Rain Portrait — Cinematic Noir") 2. The full prompt text, ready to paste into ERNIE Image 3. One line: Prompt Enhancer ON or OFF, with a one-sentence reason --- Begin.