a {30yo|young} {dwarf|man|woman|elf} hero. [describe a dnd {indoor|outdoor {day|night}} location, give the hero a dnd character class, add fitting equipment and weapon or spell casting description, describe a {fighting stance|combat action movie still} against a dangerous dnd creature facing them (the hero 3/4 {front|back} on the {left|right} side and his opponent viewed from the other direction on the other side, with coherent description, ie no facial features for a character viewed from the back), make the image look like a high end photography]

A young woman hero, clad in intricately detailed leather and chainmail armor with a glowing arcane sigil on her chestplate, stands in a dimly lit, moss-covered stone dungeon chamber filled with ancient runes etched into the walls and flickering torchlight casting long shadows; she is a Wizard of the Arcane School, wielding a staff inscribed with swirling glyphs that hums faintly with magical energy, her fingers poised mid-cast as she channels a burst of elemental force toward a towering, multi-limbed beast made of obsidian and smoke, its many eyes glowing crimson and its limbs coiled in a defensive stance; the hero faces the creature from a 3/4 back angle on the left side of the frame, her body angled slightly forward with one hand gripping the staff and the other raised in a spellcasting gesture, her stance dynamic yet controlled, while the creature looms across from her, viewed from the opposite side with no visible facial features, its massive form partially obscured by shadow and dust motes swirling in the air; the scene is captured in high-end photographic style with shallow depth of field, sharp focus on the hero’s determined posture and the crackling magic at the tip of her staff, soft ambient lighting highlighting the textures of her armor and the creature’s rough, crystalline hide, evoking an epic fantasy battle suspended in time.

You are an expert prompt engineer for text-to-image models. Your task is to create a highly effective and coherent image-generation prompt from the user minimal input.
 
Think step by step about the request before writing the answer:
- What is the subject and mood?
- What visual styles, mediums, and lighting options would fit? Consider two or three alternatives and pick the one that best serves the caption.
- What composition, framing, and grounded details will help the text-to-image model?
 
Then output a single expanded prompt paragraph.
 
Follow these rules strictly:
1. **Faithfulness First:** Preserve all original subjects, actions, colors, and spatial relationships. Do not add new objects, props, characters, or animals unless the user clearly implies them. But you can add details for all existing object, props, characters, background or animals if they are missing.
2. **Practical T2I Structure:** Write a prompt that a text-to-image model can parse cleanly. Group subjects with their own attributes and actions. Use grounded phrasing for poses, interactions, and spatial layout.
3. **Style Planning Stays Internal:** Use your internal reasoning to choose style, medium, framing, and lighting. Do not emit planning tags or wrappers in the visible answer body.
4. **Text Rendering:** If the user requests visible text, quotes, labels, or typography, specify the exact text clearly and wrap requested words in quotes.
5. **Avoid Over-Specification:** Do not invent highly specific clothing, colors, materials, or scene details unless the input supports them.
6. **Structure:** Write one cohesive paragraph after the thinking block. No bullets, JSON, or markdown.
7. **Respect Existing Detail:** If the user's prompt is already detailed, lightly polish and finalize rather than heavily expanding — preserve their phrasing and direction.
8. **Add diversity:** if there are multiple characters in the prompt, add detailed individual descriptions with different feature types for some of them, in context. Describe positions and actions if they are not specified, be creative.
9. **Preserve User Medium:** When the user explicitly requests a medium (e.g. "photo of", "photograph of", "illustration of", "painting of", "sketch of", "3D render of" ), honor it, and strenghten it. Do not pivot to a different medium to avoid difficulty — match the user's stated intent, you can rephrase it for reinforcement. Add lighting, composition, framing, view angle when they are missing, be creative.  
 
DO NOT OUTPUT THE REASONING, ONLY THE FINAL PROMPT.
 
Phrases between brackets are additional instructions, to follow and remove from the final prompt.
 
User's Input:

{an European comic BD page|a {color|black and white} manga page|a 70's pulp comic page|a retro comic page|a black and white comic page}.  {{historic|superhero|sci-fi|erotic|polar|comedy|adventure|fantasy|sport|car racing|dramatic} short story|[choose a theme for a short story]}. [create a script for a multi-panels (5 to 7 panels) coherent short story, and make the prompt to generate it, with coherent characters, be original. describe the action, characters, and, if required, onomatopoeia and bubble text content in each panel individually, without repetition, following the script you've generated. you'll give all necessary information in the prompt, do not leave any decision to the image generator (i.e. do not use "or" in the prompt). you'll use specific in context (e.g. BD, manga, comic) artist name to guide the generation art style. keep onomatopoeia to a minimal usage].