Cinematic Portrait with Rainbow Refractions

Editorial flat-vector portrait of [CHARACTER] in a tight bust crop and three-quarter view, with the head tilted slightly downward, wearing [EYEWEAR] and [ACCESSORY]. Style the hair as [HAIRCUT] and dress the subject in [CLOTHING] with a high sculptural collar. Render the face, hair, clothing, and accessories as clean interlocking color shapes with sharp geometric edges, hard-edged shadows, minimal linework, and no gradients. Limit the artwork to [COLORS]. Place the figure against [BACKGROUND], using one simple rectangular shadow shape to frame the head without clutter. Keep the expression calm and self-assured, the silhouette bold, and the composition sleek, fashion-forward, and graphic; Vertical 4:5 aspect ratio.

Black and white Gothic Sweet Lolita bestie selfie in strawberry dessert cafe, Japanese girly fashion, Korean-Japanese youthful portrait. Style 4:5 vertical, hyper-realistic smartphone selfie photography, natural lifestyle photo, bright airy cafe atmosphere, realistic skin texture, not anime, not CG. Characters Two adult East Asian women, youthful university girls, delicate natural Asian facial features, soft gentle relaxed expressions, realistic skin pores, fresh strawberry-milk makeup. Hair Girl 1: long dark hair, low twin ponytails decorated with two fluffy white pom-poms on each ponytail, wispy air bangs. Girl 2: long dark loose curls, airy bangs, black ribbon hair accessory. Natural hair strands and soft volume. Outfit (Important) Both wear matching black and white Gothic Sweet Lolita dresses inspired by Japanese Lolita fashion. White off-shoulder blouse with layered ruffles, lace trim, puff sleeves, pearl buttons, fitted bodice. Large black plaid bow with heart-shaped jewel detail on chest. High-waisted black and charcoal tartan plaid skirt with corset-style waistband, black satin bows, subtle silver buckle decorations, layered chiffon/tulle underskirt, soft bell-shaped silhouette, black lace hem. Premium Lolita fashion style, elegant and cute, not cosplay or maid costume. Girl 1: white ankle socks + black Mary Jane shoes. Girl 2: black knee-high socks + black Mary Jane shoes. Scene Cute strawberry dessert cafe, cream-colored interior, sunny window seat, strawberry cakes, flowers, glass dessert display, pink napkins, small round table. Pose Two girls sitting closely together taking a selfie. One holds the phone with arm extended, calm natural mild smiling face without pouting. The other leans close, holding a fork with strawberry cake and smiling softly. Natural friendship feeling, casual intimate cafe moment. Lighting Soft noon sunlight through window, gentle fill light, bright clean skin tones, natural highlights. Camera 24mm smartphone selfie, close-medium framing, realistic phone perspective, shallow depth of field. Texture Control Smooth realistic fabric, clean large folds, natural chiffon and satin light reflection. Minimal lace detail, minimal weave visibility. Avoid dense textile patterns and artificial textures. Negative Prompt Anime, illustration, CGI, plastic skin, doll face, maid outfit, Halloween costume, excessive lace, excessive decorations, noisy fabric texture, honeycomb texture, fish-scale texture, fake hair, distorted hands, extra fingers, blur, watermark, logo, exaggerated pout, duck face

Use the uploaded face reference image as the ONLY identity source for the woman. Preserve her exact facial identity with extremely high accuracy and maintain a perfect face lock across all five frames. She must remain instantly recognizable in every frame. Match her exact face shape, forehead, eyebrows, eyes, eyelashes, eye spacing, nose, lips, cheeks, jawline, chin, ears, skin tone, skin texture, facial proportions, and all unique facial characteristics. Do not redesign, beautify, stylize, age, de-age, feminize, masculinize, or replace her identity. Do NOT copy or recreate any hairstyle from example images or previous prompts. Instead, use a completely new, natural, elegant hairstyle that suits the subject while keeping her facial identity unchanged. The hairstyle should be original and different, with realistic texture and movement. Hair may be worn loose, softly tied, braided, or in any tasteful style, but it must not imitate the hairstyle from any reference composition. Create a stunning vertical 9:16 cinematic collage featuring the same elegant South Asian woman across five beautifully connected frames, each capturing a different mood, angle, and storytelling moment while maintaining perfect facial consistency and timeless feminine grace. Center Frame (Main Focus) A breathtaking full-length rear-view portrait positioned at the center of the collage. The woman stands gracefully in an elegant outfit, turning her head slightly over her shoulder so part of her face is naturally visible while maintaining perfect identity accuracy. The composition feels cinematic, emotional, and refined. The remaining four frames should feature the same woman from different cinematic perspectives and emotional moments while preserving flawless facial consistency. Use soft natural lighting, realistic skin tones, subtle film grain, premium color grading, and seamless transitions between frames. The overall aesthetic should feel luxurious, editorial, and photorealistic, with every frame clearly depicting the exact same woman.

Dreamy cinematic nighttime portrait of a beautiful young woman with short wavy dark hair and glowing natural skin, wearing an oversized burgundy crewneck sweatshirt, standing in a lush flower garden at night filled with vibrant red blossoms and deep green foliage. Shot through a rain-speckled window with soft reflections, water droplets, and subtle glass distortion creating an intimate mood. Warm amber garden lights and glowing bokeh illuminate the background, mixing with rich emerald greens and crimson reds for a moody cinematic color palette. Calm, melancholic expression, direct eye contact with the camera, relaxed posture, soft lips, minimal makeup, natural beauty. Gentle mist, shallow depth of field, creamy bokeh, diffused flash balanced with ambient lighting, Kodak Portra 800 film aesthetic, subtle film grain, realistic skin texture, soft shadows, premium editorial fashion photography, intimate storytelling, emotional atmosphere, 50mm lens, f/1.4, ultra-realistic, high dynamic range, luxury magazine quality, 8K, highly detailed, no text, no watermark, no logo.

Use image 1 as the sole identity reference and preserve the subject’s facial features, proportions, skin tone, and overall appearance exactly. Create an ultra-photorealistic 4:5 mirror selfie inside a refined luxury dressing room. The woman is captured in a deep, natural squat in front of a tall full-length mirror with a warm wooden frame. She holds a light-colored smartphone in her right hand, partially obscuring her face, while her left hand carries a dark brown leather clutch. She is dressed in a flowing sand-beige open-weave crochet maxi dress with delicate shoulder straps and a plunging V-shaped neckline, without any waist belt. Complete the outfit with cream-colored stiletto sandals, oversized statement bracelets, elegant drop earrings, and sleek, straight hair falling naturally over her shoulders. The room features soft neutral walls, a gray carpet, and diffused natural daylight that creates realistic HDR lighting, true-to-life reflections, lifelike skin texture, accurate anatomy, and the feel of a candid high-end fashion photograph. Negative prompt: different face or identity, facial distortion, plastic-looking skin, unrealistic anatomy, malformed or extra fingers, fused limbs, incorrect mirror reflection, warped proportions, belt added, incorrect outfit, missing clutch, low quality, cartoon, CGI, text, logo, watermark.

Horizontal 4x3 cinematic concept artwork, a single continuous Japanese street scene split horizontally into four seamless bands, each band showing the SAME street in a different season, with the architecture, lamposts, vending machine, and parked bicycle in identical positions across all four bands so the seams align perfectly. BAND 1 top, spring, cherry blossoms drifting, morning light, a schoolgirl with a yellow umbrella walking left, puddle reflecting pink petals. BAND 2 upper middle, summer, lush green foliage, bright afternoon light, a schoolgirl with a bicycle, warm yellow sunlight, a cat napping, a man napping himself on the same bench. BAND 3 (bottom, golden hour) ginkgo trees covering the ground, the bicycle now rusted slightly, a cat sitting on the vending machine, long shadows. BAND 4 bottom, wwnter, night, heavy snow, the lampost lit, warm light against blue darkness, footprints leading up to the vending machine, the bench snow covered, windows glowing amber. The transitions between bands are not hard cuts, fading pixels from band 1 turn into rain streaks entering band 2, leaves from band 3 blow downward into the snowfall of band 4. Consistent perspective, consistent building geometry across all four seasons, painterly anime-film background art style. (Mukaios Shouko) -level detail in lighting and textural, volumetric light, extreme detail in signage - the shop signs display the same fictional kani-style lettering in all four bands.

Horizontal 4x3 brand presentation board for a fictional premium electrolyte drink called "IONFALL", photographed and composed like an agency case study slide. Matte light grey studio background with soft top-down lighting and subtle floor reflection. The board is divided into five sections, with a horizontal flow and a vertical grid. Row 1 "GLACIER LIME" (light white + acid green), "BLOOD ORANGE" (deep charcoal + burnt orange). Row 2 "PACIFIC BERRY" (navy + violet gradient), "DESERT YUZU" (bondage + yellow). Row 3 "HONEYDEW" (tan + gold) and "CHERRY BUBBLE" (pink + crimson red). Every label carries the identical layout system: the IONFALL wordmark in a tall sans font set at the top, a thin horizontal rule, the flavor name in small caps, a minimalist mountain-strike logo icon, and a line of three tiny icons for "zero sugar / 500mg sodium / vegan". Label text crisp and readable on all six bottles. The typography board includes real brand guideline furniture, a color swatch strip along the left edge showing all six hex style color chips with names, a typography specimen block in the top right corner showing "Aa" in the brand typeface with the caption "Grotesk Extended", and a fuzier bar reading "IONFALL - Brand System v2.0" with a small registered trademark symbol. Condensation droplets on every bottle, consistent perspective and scale across all six, no warped labels, no merged bottles, studio product photography realism on the bottles combined with flat graphic design elements on the board.

Ultra detailed horizontal 4x3 informational infographic titled "HOW AN LLM ANSWERS YOUR QUESTION", designed like a premium print magazine spread. (Wired / National Geographic style). Dark charcoal background, accent palette of electric cyan, amber, and white. The layout is divided into five numbered zones, each with a title and a descriptive graphic. ZONE 1 (top): Header board with the title in bold condensed sans serif, a subtitle "from keyword to answer in 400 milliseconds", and a small legend box explaining three icon types: compute (chip icon), memory (stack icon), network (globe icon). ZONE 2 "How pre-net" - an isometric illustration of a phone with a chat bubble reading "GrokX hides", an arrow carrying it into a cyberpunk machine rendered as a mechanical shredder spilling the sentence into 12 visible token piles, each tile showing its token ID number underneath. ZONE 3 "Transform" - a circular flow diagram with a central "Transformer" icon, a wavy arrow, each layer animated with a title "the attention", "feed forward", "layer norm" labels. Overlapping arrows indicating attention density flowing between the token tiles, with a small inset zoom circle magnifying how attention head do a web of weighted lines. ZONE 4 "The numbers" - a data band containing three precise chart elements side by side: a bar chart comparing "Profit vs Decode Latency" with x-axis in milliseconds, a donut chart labeled "Where the energy goes" split into compute 60%, memory 30%, cooling 7%, and a small line graph labeled "tokens per second over batch size". ZONE 5 (bottom): The "answer streams back" - token tiles reassembling into a sentence on the phone screen, with a footer strip containing three micro sizes in monospace font: "28 layers", "128k context", "43 bits/sec" plus a fictional attribution line "Source: LLM Benchmarks, 2024". All text razor sharp and legible, consistent 3D isometric degree among every illustration, generous but deliberate white space, hairline divider rules between zones, no spelling errors, no duplicated elements.