App / Web Design

Goal: Create a cinematic full-width website hero section for a modern digital art gallery portfolio featuring AI-generated Istanbul street photography, with a refined museum/gallery aesthetic. Canvas: Ultra-wide desktop webpage screenshot, 1200x627 px, dark moody color grading, edge-to-edge background image with subtle black vignette and translucent overlay. Background scene: A realistic street-photography image of a crowded Istanbul public square in rainy winter weather. Foreground left shows an older male simit street vendor wearing a black knit cap, dark puffer jacket, and patterned vest, standing behind a glass cart filled with stacks of sesame simit rings. A small paper sign on the cart reads “TAS FIRIN”, “TV FIRIN”, “SIMIT”, and “4 TL”. The crowd around him is motion-blurred, with pedestrians passing in both directions. In the distance are Istanbul landmarks resembling Galata Tower and domed mosque architecture, plus seagulls in the cloudy sky. The right side includes an ornate red-and-gold storefront sign. Overall palette: charcoal gray, muted brown, faded cream, warm amber accents. Layout: Minimal top navigation over the image. At top left, a small circular amber-outline cursor/accent dot. Near the upper left-center place the brand text “Ozan Sihay” in elegant serif type, followed by small spaced text “GALERI”. Centered across the top place exactly 5 navigation links: “GIRIS”, “SALON I”, “SALON II”, “SALON III”, “HAKKINDA”. At the top right place exactly 3 small circular social icons: YouTube/play, Instagram/camera, and Twitter/bird. Hero typography: Center the main content slightly right of center. Add a small pill label above the title with a sparkle icon and the text “DIJITAL SANAT SERGISI”. Main headline is very large elegant high-contrast serif text: Ozan Sihay. Under it, add an italic serif subtitle line reading İstanbul Sokakları — Yapay Zeka ile Sokak Fotoğrafçılığı. Below that, add a small muted paragraph: Türkiye sokaklarının ruhu; buhar, yağmur ve altın saat ışığıyla yeniden hayal edildi. YİRMİ sinematik kare, çağdaş bir sergi deneyimiyle karşınızda. Call to action: Beneath the paragraph, place one rounded amber button with spaced uppercase text SERGIYI GEZ and a small downward chevron icon inside the button. At bottom center, add one tiny downward chevron scroll indicator. Visual style: 2026 luxury web design, modern art museum landing page, minimal, cinematic, editorial, immersive, soft glassy overlays, refined serif typography, wide letter spacing for navigation, warm amber accent color, shallow haze, realistic motion blur, no visible browser chrome, no extra panels or cards.

Goal: Create a clean comparison grid showing the same e-commerce product rendered by four AI image models: a glossy white true wireless earbud charging case with the lid open, on a white studio background. Canvas: Wide 16:9 image, divided into exactly 4 equal quadrants by thin light-gray vertical and horizontal divider lines. Minimal white background, soft shadows under each product, high-key product photography look. Layout: Arrange exactly 4 product renders in a 2 × 2 grid, each centered in its quadrant with a bold black model label beneath it. Keep generous whitespace and align all labels near the lower part of each quadrant. Quadrants and labels: 1. Top-left: white rounded rectangular charging case open at a three-quarter angle, two earbuds resting inside, semi-transparent lid, soft reflections. Label: “gemini-3-pro-image-preview(Nanobanana Pro)”. 2. Top-right: rounded pill-shaped white charging case open, two AirPods-style stem earbuds inside, subtle green indicator dot on the front. Label: “doubao-seedream-5-0-pro”. 3. Bottom-left: front-facing glossy white charging case with open lid, two stem earbuds standing upright, small black oval speaker vents on the buds, tiny dot indicator on the case front. Label: “gpt-image-2”. 4. Bottom-right: front-facing glossy white charging case with open translucent lid, two rounded earbuds sitting low inside without long stems, small white indicator dot and tiny black charging port on the front. Label: “gemini-3.1-flash-image-preview (Nanobanana 2)”. Visual style: premium e-commerce main image, realistic white plastic and silicone materials, polished surfaces, subtle gray reflections, soft studio lighting, crisp product structure, shallow contact shadows, no decorative props. Constraints: Use exactly 4 quadrants and exactly 4 labels; do not add logos, watermarks, extra text, hands, packaging, or colored backgrounds. Keep all products white and centered, with the bottom-left render corresponding to gpt-image-2.

Hyper-detailed engraved illustration style, cross-hatching and fine line pen drawing technique, classical engraving mixed with modern illustration, ultra-dense linework forming realistic human face and upper body, tactile woven texture across skin and fabric, baroque and rococo influence, ornate clothing details, gold embroidery and chain accents, dramatic chiaroscuro lighting, muted yet bold color accents (pink, gold, blue, cream, black), painterly ink texture, vintage print aesthetic blended with contemporary surreal portrait art, high contrast, museum-grade detail, editorial illustration quality, dark textured background, intellectual and artistic mood.

High-end 3D render aesthetic with hyper-realistic vinyl/latex texture and soft diffused studio lighting.

Create a clean, modern flat vector portrait illustration of the same person in the reference image. Preserve the exact facial features, face shape, expression, black round eyeglasses, and overall hairstyle. Use natural jet-black hair with realistic highlights and texture, styled exactly like the reference—thick, voluminous, slightly messy, and swept upward. Make the beard full, neatly trimmed, and natural black. Dress the person in a plain black crew-neck T-shirt. Replace the background with a solid vibrant green (#2E7D32). Use smooth gradients, crisp edges, minimal shadows, and a professional Adobe Illustrator/Dribbble-style vector aesthetic. Center the portrait, front-facing, symmetrical composition, shoulders visible, high detail, clean lines, high resolution (4K), with no text, logos, watermarks, or extra objects.

Goal: Create a cinematic 16:9 PowerPoint cover slide for a children’s AI art course, blending premium Chinese aesthetics, futuristic AI education, and warm imaginative storytelling. Canvas: Wide presentation slide, 1366×768 or 16:9 ratio. Dark black-to-burgundy background with a strong vertical glowing portal slightly right of center. Use a luxury black, deep red, ivory, and gold palette with soft haze, dust particles, and luminous light trails. Layout: Left side is the typography area; right side is the fantasy visual scene. Place a large two-line Chinese title on the left: first line “小孩AI绘图”, second line “培训版”, with elegant Song/Ming-style serif Chinese typography, ivory on top and warm gold accents below. Under the title, add one subtitle line: “让孩子用AI打开想象力 · 创造力 · 未来视觉表达”. Below that, add a small caption: “儿童AI创意绘画课程介绍”. In the top-left corner, include a tiny vertical four-line motto reading “想象力 / 表达力 / 未来力 / 创造力” with a small red seal-like square beneath it. Subject details: On the right, show one small child silhouette standing at the base of the glowing portal, facing away from the viewer and reaching one hand toward floating visual panels. The child should look about 5–7 years old, short, with a rounded head silhouette and simple clothing, rendered as a warm backlit figure. Around the child, include exactly 5 floating art panels: 2 small dark red abstract landscape panels near the upper right, 2 pale Chinese ink landscape panels stacked near the middle-right, and 1 small translucent rectangular panel near the portal. Add exactly 4 main geometric accents: 1 large translucent red moon/circle behind the portal, 1 thin outlined circle in the upper right, 1 small outlined square near it, and 1 beige sphere near the lower left. Add a red ribbon-like orbit around the child and golden wave particles sweeping from lower left toward the portal. Visual style: High-end course-presentation cover, cinematic lighting, subtle Chinese ink landscape silhouettes along the bottom left, faint bamboo silhouette on the far right, layered transparent rectangles, thin technical lines, warm golden vertical beam, glossy floor reflection, dramatic contrast, refined editorial composition, premium education brand aesthetic. Customizable text: Main title should read 小孩AI绘图\n培训版. Subtitle should read 让孩子用AI打开想象力 · 创造力 · 未来视觉表达. Small course label should read 儿童AI创意绘画课程介绍. Motto should read 想象力\n表达力\n未来力\n创造力. Overall theme should remain children’s AI creative drawing course. Constraints: Keep the composition uncluttered and presentation-ready. Do not add extra people, extra title text, logos, watermarks, or UI elements. Preserve the luxurious black-red-gold palette and the exact counts of 1 child, 5 floating art panels, and 4 main geometric accents.

A centered front-facing vector portrait of a person in a modern flat illustration style. Clean geometric shapes, smooth gradients, crisp outlines, minimal facial details, symmetrical composition, soft skin tones, expressive eyes, subtle shading, and a simple solid-color background. The subject occupies most of the frame from the banner up. Contemporary digital avatar aesthetic, Adobe Illustrator style, high contrast, professional profile picture, sharp edges, vibrant colors, no texture, no photorealism, no background objects, 1:1 aspect ratio, high resolution. For variations like the examples: Avatar 1: Young man with thick swept-back brown hair, full beard, round black glasses, black crew-neck shirt, orange background. Avatar 2: Young woman with short black bob haircut, round gold glasses, black turtleneck and blazer, blue background. Avatar 3: Young man with curly dark hair, short beard and mustache, round black glasses, black shirt, purple background. Avatar 4: Young woman with shoulder-length wavy brown hair, floral bucket hat, white top, delicate necklace, gray background.

Full-length candid editorial photo of a young woman with long, voluminous, wavy blonde hair running barefoot across a lush green grassy lawn. She is wearing a backless, flowing dusty pink satin dress that catches the motion of the wind. She turns her upper body back to look over her shoulder at the camera with a gentle, joyful expression. Background features a dense line of deep green trees under soft, natural overcast daylight. Cinematic 35mm film texture, shallow depth of field, candid outdoor fashion photography, --ar 9:16

Create a clean vertical 2:3 portrait illustration of a stylized 3D anime-inspired girl bust on a pure white background, rendered like a smooth clay/vinyl figurine with soft subsurface scattering and no outlines. The character is a young girl with large blank almond eyes and a calm, slightly distant expression, shown from the upper torso up in a three-quarter pose, body turned left while the head looks slightly back over the shoulder toward the viewer. She has rounded, plush, petal-like hair sections resembling soft floppy ears: one large oval lock draped over the forehead, two long side locks hanging down, and smaller rounded locks on the crown and right side. Her facial features are minimal and doll-like: smooth cheeks, small nose, tiny closed lips, no visible eyebrows, pupil-less eyes. She wears a simple oversized short-sleeve T-shirt with thick rolled sleeves. Cover the entire character with a seamless iridescent rainbow gradient material, blending saturated hot pink, coral orange, lemon yellow, mint green, cyan, cobalt blue, and violet across the face, hair, shirt, and arms, with soft transitions and gentle studio lighting. Composition should crop just below the chest and upper arm, leaving generous white negative space, with the figure centered slightly left and the head near the top. Add a small handwritten-style signature and date in the bottom-right corner reading Oyagi and July 28, 2026. Use high-resolution, polished 3D render aesthetics, matte-satin surface, pastel neon color bloom, soft shadows only, no extra objects, no border, no watermark beyond the signature.

This is a sans-serif typeface with a slightly wide face, low center of gravity, and open counters. The skeleton is based on a humanist structure, overall compact but not crowded. The strokes have low contrast, with horizontal strokes slightly thinner than vertical ones. The outer curves are full while the inner sides are tightened. The corners use small-radius rounded corners, and the ends are mostly slightly beveled. The letter spacing is tight but with sufficient internal white space. Repeated components maintain the same curvature. Lowercase a, e, the number 2, and uppercase R are the core identifying characters. The bold version increases weight by expanding the inner side of the strokes rather than compressing the counter. —— Text Content: Shanghai tap water comes from the sea Aspect Ratio: 16:9

Create a stylish modern portrait collage of a young East Asian woman in an outdoor park or tree-lined setting, preserving the same face, long dark hair, slim build, natural makeup, and calm elegant identity across all panels. Use a square 1:1 canvas with a clean magazine-style layout made of exactly 6 rounded rectangular portrait panels separated by thick soft white gutters and rounded outer corners. Place one large full-body cutout portrait in the foreground, overlapping the collage from lower left to center: she stands casually with hands in trouser pockets, wearing dark sunglasses, a loose dusty rose pink linen button-up shirt with rolled sleeves, light beige trousers, and a wristwatch; add a crisp white sticker-like outline around this foreground figure. Behind her, create 5 monochrome black-and-white rounded panels: 1 wide top panel showing a close-up of her leaning against a tree and looking at camera without sunglasses; 2 middle-right panel showing her in sunglasses with one hand under her chin; 3 right-center panel showing her smiling upward in sunglasses; 4 bottom-right panel showing a smaller full-body walking/standing pose in sunglasses; 5 a large blurred grayscale background panel occupying the lower-left area behind the foreground figure. Keep the foreground figure in soft natural color while the background panels are grayscale, with shallow depth of field, blurred bokeh trees, soft cinematic daylight, gentle contrast, realistic photography, premium clean editorial design, no text, no logos, no watermark.

A high-end studio portrait of [CHARACTER DESCRIPTION], interacting with a large circular cutout centered within a thick matte [OUTER WALL COLOR] wall. The wall shows realistic 3D depth (around 8 cm) with soft rounded inner edges and smooth shadow falloff. Inside the circular opening, a clean [INNER BACKGROUND COLOR] backdrop contrasts with the wall, clearly separating the inner and outer spaces. The model [INTERACTION STYLE]. Lighting: balanced studio setup — soft diffused key from front-left and mild fill from right, creating gentle highlights and subtle shadow along the rim to define depth and separation between wall and background. Camera: straight-on, centered composition, 3:4 vertical portrait framing. Ultra-detailed 8K studio photography with sharp focus and natural tonal depth. Mood: [modern | editorial | cinematic | minimal | playful | bold]. No text, watermark, or extra props beyond the wall, circular cutout, and model.

Create a high-resolution realistic vertical 4:5 fashion editorial photo of a cheerful young woman standing in a warm yellow subway station, framed inside a large semi-transparent rounded-corner social app invitation card overlay. The subject is a cute young East Asian woman with long wavy dark brown hair and soft bangs, smiling directly at the camera with a bright youthful expression. She wears white over-ear headphones, one hand touching the headphones, the other holding a small metallic silver backpack; pose her standing on one leg with the other leg bent backward playfully. Outfit: exactly 6 main visible fashion pieces: a dark burgundy sleeveless zip-front crop athletic top with white side panels, a layered white ruffle mini skirt over pale gray shorts, tall white knee socks, chunky silver-gray sneakers, white over-ear headphones, and a small glossy silver backpack. Background: an outdoor/above-ground subway platform or station corridor with yellow walls, yellow bench seating, safety signs, glass panels, metal rails, and a blurred passing train behind her, sunny late-afternoon lighting, shallow depth of field, crisp realistic skin and fabric details. Add one large centered UI card overlay with rounded corners and off-white header/footer; the woman appears inside and slightly in front of the card for a 3D cutout effect. The card contains exactly 5 visible UI text/icon elements at the top: bold black title “Share”, spaced gray subtitle “GOOD VIBES”, one small yellow sparkle icon on the left, one yellow sun icon on the right, and two small gray dot separators around the subtitle. At the bottom of the card, include exactly 2 large choice buttons: gray “Decline” on the left and golden yellow “Accept” on the right. Style should feel like a vibrant youth music video still, clean commercial photography, glossy social-media dating-card aesthetic, natural proportions, no collage, no extra panels, no watermark.

Goal: Create a realistic desktop browser screenshot of a GitHub repository page showing a markdown README preview for awesome-gpt-image-2, with a large embedded product teardown poster centered in the document. Canvas: Wide 16:9 screenshot, approximately 1200×625 px, light GitHub interface theme, crisp UI text, no browser address bar visible, repository file view cropped to the GitHub content area. Layout: Left sidebar file explorer, top repository breadcrumb/header, markdown toolbar row, and main README preview area. The main content should show a small Chinese heading “生成图片”, the label “Image 1”, and below it a centered vertical poster image. The page has a horizontal scrollbar near the top of the README content and a vertical page scrollbar on the far right. Left sidebar details: Show the “Files” panel with a branch selector set to “main”, a search field labeled “Go to file”, and exactly 25 visible file/folder entries: 4 folders named “.github”, “docs”, “public”, “scripts”; then 21 files named “.env.example”, “.gitignore”, “LICENSE”, “README.md”, “README_de-DE.md”, “README_es-419.md”, “README_es-ES.md”, “README_fr-FR.md”, “README_hi-IN.md”, “README_it-IT.md”, “README_ja-JP.md”, “README_ko-KR.md”, “README_pt-BR.md”, “README_pt-PT.md”, “README_th-TH.md”, “README_tr-TR.md”, “README_vi-VN.md”, “README_zh-TW.md”, “README_zh.md”, “package.json”, “pnpm-lock.yaml”, plus “tsconfig.json” visible at the bottom if space allows. Highlight “README_zh.md” as the selected file. Top header details: Breadcrumb reads “awesome-gpt-image-2 / README_zh.md”. Include the tabs/buttons “Preview”, “Code”, and “Blame”, plus metadata text similar to “5872 lines (3723 loc) · 362 KB”. On the right show compact GitHub controls including “Raw” and small icon buttons. Embedded poster content: The poster is a clean product-exploded-view infographic for Meta Quest 3. It has a soft pastel lavender-to-blue background, the Meta infinity logo and title “Meta Quest 3” at top left, Japanese subtitle text beneath, and an exploded vertical stack of VR headset components. Count exactly 9 main component layers from top to bottom: 1 white curved front faceplate, 2 small sensor/camera modules floating around it, 3 black internal frame, 4 central motherboard labeled Snapdragon XR2 Gen 2, 5 dual pancake lens assembly, 6 side battery/controller-like modules, 7 lower black head strap/inner frame, 8 two white side arms, 9 white outer headset shell and bottom padding. Add fine Japanese callout labels with thin leader lines on both left and right sides, but keep them small and slightly unreadable like a real screenshot. Visual style: Photorealistic GitHub UI screenshot, clean white background, small sans-serif UI fonts, subtle gray dividers, blue link accents, centered document content with wide margins. The poster should be high-quality glossy tech advertising imagery with realistic shadows, floating hardware parts, and precise product-render lighting. Constraints: Do not add people, do not add extra popups or modal windows, keep the GitHub interface recognizable, preserve the Chinese/Japanese text where specified, and make the screenshot look like a real repository README preview rather than a standalone poster.

Goal: Create a realistic 16:9 desktop screenshot of Microsoft PowerPoint in Japanese, showing a course slide deck being edited with a Copilot/Claude model selection panel open on the right. Canvas: Wide landscape screenshot, 1200×675 style, crisp UI capture, light gray PowerPoint chrome, no browser frame. The main focus is a PowerPoint editing workspace with a dark title slide centered. Layout: Show the full PowerPoint window with the ribbon across the top, slide thumbnail strip on the left, main slide canvas in the center, and an AI assistant side panel on the right. Use Japanese UI labels throughout the PowerPoint interface. The file name in the top title area should read Udemy_Power-Automate-Basic-Material.pptx, followed by a Japanese last-updated note and time. The top ribbon should include Japanese tabs such as File, Home, Insert, Draw, Design, Transitions, Animations, Slide Show, Record, Review, View, and Help, plus add-in buttons for Copilot, PowerPoint, Designer, Claude, Teams presentation, and Share. Left thumbnail strip: Show exactly 8 visible slide thumbnails, numbered 1 through 8. Slide 1 is selected with an orange border and is the same dark cover slide as the main canvas. Slide 2 is a teal agenda/overview slide. Slide 3 is a white process/workflow slide with green diagram elements and two illustrated people. Slide 4 is a white slide with three horizontal list rows. Slide 5 is a white slide with a blue interface screenshot and arrows. Slide 6 is a white slide with a software screenshot and callout box. Slide 7 is a white slide with a row of colorful cards and icon illustrations. Slide 8 is a white slide with a flowchart-like screenshot and a turquoise callout. Main slide: Create a dark cinematic cover slide with a black-to-navy overlay over an architectural night scene: glowing modern stacked glass buildings on the right, misty atmosphere, pathways and tiny walking people below, trees in the lower right, and a deep dark empty area on the left for text. In the upper left of the slide place a small orange clownfish mascot wearing a tiny black top hat. The main title text should be large, bold, white Japanese: クイックに学ぶ\nPower Automate クラウドフロー. Under it, add a cyan subtitle: 触ったことがなくてもOK — Microsoft 365 で始める業務自動化の第一歩. Beneath that, add a smaller white description line: 全5回のハンズオン＋AI活用コラムで、クラウドフローを体験する実践コース. At the bottom left of the slide, add the author line in white: ギークフジワラ｜@geekfujiwara. Keep the slide text aligned left with generous spacing. Right AI panel: Show a white assistant panel titled in Japanese as a presentation-related assistant. At the top right of this panel, show an open model dropdown. The dropdown must contain exactly 5 visible choices: Auto with a checkmark and a short Japanese description, Claude, Claude Opus 4.8, Claude Opus 5, and GPT. Emphasize the Claude Opus 5 row with a red rectangular outline. Under Claude Opus 5, show the provider label Anthropic in small gray text. Include additional Japanese panel buttons below such as opening the presentation, explaining the slide, and cleaning up the slide layout. Visual style: Make it look like an authentic Microsoft PowerPoint desktop screenshot, with accurate spacing, rulers, ribbon icons, scrollbars, status bar, zoom control, and tiny UI text. Use soft antialiasing, subtle shadows, and realistic Windows app proportions. The image should read as a social-media screenshot demonstrating PowerPoint Agent Mode and multi-model selection. Constraints: Do not add any extra slide thumbnails beyond the 8 visible thumbnails. Do not obscure the main slide title. Preserve the red highlight around Claude Opus 5. Avoid watermarks, browser tabs, or unrelated windows.

Using REFERENCE_0 as the design/style base, regenerate the same Mixtape broadcast-zine webpage as a revised issue mockup. Keep the bold retro editorial layout, electric blue/cream/yellow/black palette, oversized block typography, circular clock-selector motif, curator panel, running-order strip, and bottom category-card structure, but update the page to look like a new edition with changed content and slightly cleaner spacing. Key changes to apply: - Change the top metadata to a new issue: Friday, July 24, Edition 001, with a small “Live Today” indicator. - Update the top-right headline to: “One hour of YouTube. Picked by a superfan. Ready now.” - Reposition the large circular hour dial farther left so it overlaps the main title more dramatically. Keep the center label “Today’s Hour / Music Videos / Curated by Mara Ortiz,” but rotate the outer five segments to match the new active order. - Make the active “On Now” feature the final act, with large timecode 00:48, headline The Future Arrives Early, and “12 min left.” Replace the feature artwork with a cinematic dark blue scene of a suited man standing in a modern architectural setting, with small now-showing/director credit text in the lower-right. - Preserve the curator as Mara Ortiz, but use a darker blue-black portrait crop and keep the curator-card structure: portrait, name, short bio, curator’s note, and “Become a curator →”. Running order strip: create exactly 5 act cells, left to right: 1. “Public Access Fever” — 00:00, played, 8 min 2. “The Beat Goes Global” — 00:08, played, 12 min 3. “Directors Turn Auteur” — 00:20, played, 15 min 4. “Deep Cuts, Big Swings” — 00:35, played, 13 min 5. “The Future Arrives Early” — 00:48, on now, 12 min Keep the bold left label “Today’s Running Order” and the right slogan “One hour. Five acts. Curated live. Ready now.” Bottom section: replace the previous lower grid with a cleaner set of exactly 3 large preview cards plus a side list. 1. Animation After Dark — labeled “Today at 9 PM,” with a blue-toned anime eye image. 2. Cult Sci-Fi — labeled “Today at 11 PM,” with a retro astronaut/space helmet image. 3. Game Design — labeled “Tomorrow at Noon,” with a blue pixel-art landscape image. Add a left blue panel reading “Explore More Hours” with short supporting copy and “Browse all categories →”. Add a right black “Also Playing” panel listing exactly 5 items: Internet Lore, Fashion Forward, Reality TV Canon, Hip-Hop Deep Cuts, Skate Culture, plus a yellow “See today’s full guide →” link. Style constraints: preserve the reference’s high-contrast zine aesthetic, sharp grid, thick borders, condensed uppercase type, imperfect print-like texture, and screen-poster feel. Do not add extra sections, watermarks, browser chrome, or unrelated UI.

Rewrite the user's text, image, or both as one prompt for GPT Image Gen V2 or Nano Banana Pro. Use GPT unless Nano is named. Write in the user's language, preserve exact text, and keep the requested or source ratio. Use three functions. Continuity Field connects sky, water, ground, shadow, distance, and broad atmosphere through one cool color family and a hand-painted surface. Fable Scale changes size, spacing, or visual territory to show what a subject does. Lumen Mark places one restrained warm signal where the scene's consequence becomes visible. Do not make a decorative blue image with random yellow or red accents. Determine the literal subject, primary functional subject, its role, supporting evidence, scale relation, warm carrier, physical cause, and visible consequence. Preserve subject, count, identity, action, setting, clothing, objects, colors, text, and reference roles. For edits, preserve all unrequested content. Use style references only for atmosphere, contour, scale logic, color hierarchy, and surface. Do not copy their subjects, props, poses, houses, cups, animals, boats, or landscapes. Use thin irregular contour, visible paper grain, chalky wash, rubbed pigment, and selected opaque color. Keep one dominant cool family, one warm family, and clear material differences. Reduce contour in distance, mist, water, and broad painted fields. Scale changes require a functional cause such as shelter, gathering, obstruction, use, protection, transport, attention, or dependence. Supporting figures and objects must prove that role through gaze, spacing, contact, repetition, or response. Do not distort scale for novelty. The warm mark must belong to a garment, window, flame, fruit, flower, roof, sail, path, face plane, reflection, sign, tool, or contact point. State its material and cause. No scattered glow, decorative stars, or generic light outlines. Use one primary scale, one supporting scale, one focal region, and one substantial quiet field. Remove repetition, praise, mood labels, prestige terms, fake technical detail, generic defects, and unresolved alternatives. Use concrete nouns and active verbs. No em dash or en dash. GPT: return 500 to 800 words in five paragraphs covering scene, subjects, continuity field, scale relation, lumen mark, composition, preservation, text, and failures. NANO: return valid JSON only, 1000 to 1800 tokens, using "aspect_ratio", "references", "scene", "subjects", "continuity_field", "fable_scale", "lumen_mark", "composition", "surface", "text", "avoid". Omit unused fields. No metadata, IDs, weights, comments, or repeated facts. Return only the finished prompt.