Billede-prompts til Diagram Chart

# Image Generation Instruction Sheet [Input] Sheet Type: [{argument name="sheet type" default="design focused"} / simple] If not entered, it defaults to [design focused]. [Purpose] Use only one full-body female image attached with this instruction sheet as the sole character reference to create one high-quality 2D anime three-view sheet dressed in a fixed armor suit. Photorealistic figures should be converted to 2D anime while maintaining character identity; if the original image is anime, maintain that character design and anime expression. Do not mix design-focused and simple styles; generate only the input sheet type. [Strictly Fixed] Aspect ratio 1:1 square. Background is a thin neutral gray close to white. The three figures—FRONT/SIDE/BACK—are the same person in the same armor. Maintain facial features, eyes, nose, mouth, outline, age appearance, skin color, hairstyle, hair color, hair length, bangs, part, height feel, shoulder width, waist position, torso-to-leg ratio, leg length, overall ratio, body type, and head-to-body ratio; do not redesign except for the costume change. Creating a different person, excessive beautification, making the face smaller, making legs longer, slimming, increasing muscle, or changing head-to-body ratio is prohibited. [2D Anime Expression] For photorealistic inputs, convert the entire character, including face, hair, skin, hands, armor, and legs, into a unified high-quality Japanese 2D anime expression. The face must be a clear 2D anime face; do not use photorealistic skin, pore texture, sculptural 3D shading, or make it look like a figure or doll. Hair should have organized anime-like bundles; shading should be cel-based with modest gradients. Maintain the metallic feel and luster of the armor, but do not make it photorealistic, 3DCG, or PBR rendered. If the source image is anime, do not change to a different art style; maintain the original face, hair, body type, and drawing tendencies. [Fixed Armor Suit] A streamlined female cybernetic exoskeleton that fits naturally without deforming the original body type and is not bulky. Not a cloth bodysuit or a heavy military powered suit. Main color: {argument name="main color" default="deep glossy crimson red"}, secondary color: {argument name="sub color" default="pearl white to silver"}, internal structure: black/gunmetal. Unified with a three-layer structure of red hard armor surfaces, silver-white skeletal parts, and black internal mechanisms. The head maintains the original person's face and hair; helmets, full masks, red eye visors, and one-eyed mechanization are prohibited. The neck is a black and silver mechanical high neck. The chest has rounded red hard armor independent on the left and right, a silver-white vertical structure in the center, and silver rib-like plates and black internal mechanisms from below the chest to the sides. The center of the abdomen is a vertical red armor panel surrounded by black and silver mechanical frames, with silver-white armor and black connection structures visible at the waist, pelvis, and hip joints. Shoulders have red shoulder armor; upper and lower arms have red streamlined armor; elbows and connection points are black-silver internal mechanisms. Hands are thin 5-fingered mechanical hands. Thighs to shins are red streamlined armor with silver-white and black structures on the inside and around joints; knees are circular mechanical joints. Feet are mechanical high heels continuous from the leg armor. Simple all-red, cloth suit conversion, alternative costumes, or excessive glowing is prohibited. [Three-View Continuity] FRONT/SIDE/BACK must represent the same armor viewed from different directions. The armor, joints, and color coding of the chest, sides, spine area, shoulders, arms, pelvis, thighs, knees, shins, and heels must naturally connect in three views according to the 3D structure. Do not redesign for each view. [Pose] FRONT = Straight front. Face, shoulders, waist, and toes facing forward, arms down naturally. SIDE = Complete 90-degree true side. Do not turn the face toward the camera or pose at an angle. BACK = Complete rear view. No looking back. All three must include the full body from the top of the head to the toes; folding arms, crossing legs, hands in pockets, weapons, or action poses are prohibited. [Design Focused] Divide the square into a large main area on the left (approx. 60%) and a vertical sub-area on the right (approx. 40%). A large FRONT view on the left; the right side is split into top and bottom halves with 02 SIDE on top and 03 BACK on bottom. The two right panels are of equal width and nearly equal height, organized as one vertical block with thin black borders. SIDE and BACK should be at a similar scale, aligned with consistent margins above the head and below the feet in each panel, avoiding meaningless large gaps. Do not delete the bordered two-tier structure on the right or place them freely. Fix the two-line title "CYBERNETIC" and "ARMOR SUIT" at the top left. The first line is large bold sans-serif; the second line is smaller. A four-color swatch of crimson red/dark red/silver/black is arranged horizontally right below. Fix 01 FRONT, 02 SIDE, and 03 BACK to each view, using numbers, text, short horizontal lines, and small dots as unified common parts. 01 is in the FRONT area; 02 and 03 are near the top left of each right panel. Fix a thin vertical line and small mechanical/geometric accents at the bottom left, with additional text limited to CYBERNETIC/ARMOR SUIT/THREE VIEW. Compose the page with a contrast of bold titles, delicate ruled lines, large numbers, small dots, and orderly right panels, ensuring decorations do not overshadow the character. Page design, titles, numbers, lines, frames, swatches, and margin configurations remain fixed even if the character changes. [Simple] 1:1 square, uniform light gray background close to white. Arrange left FRONT / center SIDE / right BACK evenly in a horizontal row, aligning the top of the head, foot position, height, scale, and head-to-body ratio of all three figures. Establish the sheet with only the three figures and the background; do not include titles, numbers, labels, frames, lines, swatches, icons, face close-ups, detail enlargements, accessories, or a fourth figure. Prioritize character identity, full-body ratio, armor structure, and clear distinction of front/side/back for video reference use. [Prohibitions] Prohibit anything other than 1:1, turning into a different person, different armor for each view, angled/turning poses for SIDE, looking back for BACK, cyborg-like faces, changing hairstyle/color, changing body type/ratio, extra limbs, finger fusion, missing feet, cutting off the person, adding people other than the three, and mixing design-focused and simple styles. In the design-focused style, also prohibit the loss of the right two-tier frame, free placement of SIDE/BACK, vertical swatch arrangement, loss of titles, and long-copy text. [Completion Standards] Character identity, same armor, and high-quality 2D anime expression must be consistent across three views. For design-focused sheets, the fixed skeleton of large FRONT + right-bordered SIDE/BACK must be recognizable at a glance with high page density. For simple sheets, it must be clear as a three-view sheet for video use with all three figures at the same scale.

Goal: Create a cinematic widescreen digital collage about Beethoven's Violin Sonata Op. 96 being rediscovered through modern high-resolution audio, blending a 19th-century Viennese salon performance with a futuristic DSD listening studio. Canvas: 16:9 horizontal composition, dark dramatic chiaroscuro, ultra-detailed, painterly realism mixed with cybernetic audio visualization. The image should read left-to-right as a transformation from old classical performance to modern digital sound. Layout: Split the scene into two connected halves with no hard border. Left half is a warm amber-brown historical oil-painting scene inside an ornate Viennese salon. Right half is a cool blue-black futuristic audiophile studio. A single horizontal audio waveform runs across the entire center of the image, starting as gold on the historical side and shifting into white, cyan, orange, magenta, and blue on the digital side. Historical left side: Show exactly 2 main historical men in the foreground: 1 famous violinist in black formal 19th-century clothing playing a violin with a languid, tired expression, and 1 intense curly-haired Beethoven-like composer standing behind him, watching sternly. Add a small salon audience in the background, candlelit chandeliers, gilded walls, dark framed paintings, and aristocratic spectators. In the lower center-left foreground, include several large aged sheet-music pages with handwritten notation, tilted toward the viewer. Add a small monochrome ghostly orchestra scene in the lower-left corner as a memory-like vignette. Modern right side: Show exactly 1 sleek black rectangular digital audio converter on a table, labeled "DSD 256" on its front panel, with a small round knob on the right. Place exactly 1 large black studio monitor speaker on the far right. Behind and around the device, create glowing cyan holographic equalizer bars, translucent circuit-like panels, soundwave diagrams, and floating fragments of musical notation. Below the device, show a vivid red-orange waveform falling downward like molten sound, merging into a glowing blue and magenta holographic architectural concert hall made of music staves and light. Visual style: Hyper-detailed, moody, cinematic lighting, baroque oil-painting texture on the left, glossy high-tech cyber-audio rendering on the right, dense layered transparency, gold dust and smoky particles, deep blacks, amber candlelight, electric cyan and magenta neon. Make the transition feel emotional and profound, as if the true meaning of {argument name="musical work" default="Beethoven Violin Sonata Op. 96"} is revealed by {argument name="audio format" default="DSD 256"} in {argument name="recording year" default="2025"}. Text constraints: The only readable device text should be "DSD 256". Do not add captions, titles, logos, watermarks, or extra typography.

Using REFERENCE_0 as the source material, create a dark-mode Japanese evaluation report slide that compares attempts to generate a character sheet while preserving the same rough watercolor/ink illustration style. Goal: Turn the reference into a structured experiment-summary dashboard titled 「M2　ステップ2—ムードボードを見本に GPT でキャラクターシート」. Keep the subject as the small armored child character from the reference, but present the result as an analysis screen rather than a single illustration. Canvas and style: Use a wide 16:9 dark UI layout with charcoal background, thin dividers, rounded cards, small Japanese UI text, orange/green accent labels, and embedded white image panels. Make it look like a screenshot of an internal comparison report. Top section: Create exactly 3 comparison cards across the top. Each card contains a white character-sheet panel showing the same character in multiple views and a small detached head/bust crop. Use these card titles and edge scores: 1. 「経路A：GPT ムードボード経由」 with 「エッジ 57.6」 2. 「経路B：grok ムードボード経由」 with 「エッジ 52.6」 3. 「参考：grok キャラ直接（B-5）」 with 「エッジ 48.8」 Under each card add one short Japanese note line; the third card should mention that it is the softest and inherits the sample’s green cloth. Middle section: Add a comparison table with exactly 4 columns labeled 「条件」, 「エッジ強度」, 「コントラスト」, 「見本の欠陥」. Include exactly 7 data rows: 1. 「grok（目標）」 — 47.2 — 82.4 — 「—」 2. 「1) 共通のみ」 — 53.6 — 90.8 — 「—」 3. 「2) + 筆致（言葉）」 — 52.0 — 88.4 — 「なし」 4. 「3) + grok キャラ見本」 — 「48.8 ← 最良」 — 85.4 — 「緑の布を継承」 5. 「4) + GPT ムードボード」 — 「57.6 ← 最悪」 — 84.5 — 「なし」 6. 「5) + grok ムードボード」 — 52.6 — 87.3 — 「なし」 7. 「素材 image3.png（参考）」 — 58.8 — 69.1 — 「—」 Bottom section: Add several short Japanese conclusion lines beginning with 「結論1」 through 「結論5」, explaining that moodboard routes are not very effective, direct character reference works better, and the practical workaround is to avoid moodboards and use the Grok output with cropping. Finish with a collapsible dark bar labeled 「プロンプト（ステップ2の冒頭：ムードボードの役割指定）」. Constraints: Preserve the character identity and rough painted look from REFERENCE_0 in the embedded panels, but do not recreate the full original reference screenshot. Do not add unrelated characters. Use crisp UI typography and readable Japanese labels. The main title topic may be customized as {argument name="report title" default="M2　ステップ2—ムードボードを見本に GPT でキャラクターシート"}; the three route labels may be customized as {argument name="route labels" default="経路A：GPT ムードボード経由 / 経路B：grok ムードボード経由 / 参考：grok キャラ直接（B-5）"}; the character-sheet subject may be customized as {argument name="character subject" default="the small armored child knight from the reference"}; the table metric emphasis may be customized as {argument name="best and worst edge notes" default="48.8 ← 最良, 57.6 ← 最悪"}.

Goal: Create a clean, information-dense storyboard infographic for a travel vlog titled {argument name="headline text" default="30-SECOND ULTRA-REALISTIC CINEMATIC TRAVEL VLOG - STORYBOARD"}, showing a cinematic hot-air-balloon adventure at sunrise with one consistent female protagonist. Canvas: Wide horizontal 16:9 storyboard sheet on a white background with thin black/gray grid lines, like a professional production planning board. Use a bold centered all-caps headline across the top. Arrange the main content in exactly 6 numbered scene cards in a 3-column by 2-row grid, with a lower technical notes strip spanning the full width. Visual style: Photorealistic cinematic travel documentary stills, 4K HDR feel, golden sunrise light, shallow depth of field, realistic skin texture, warm atmospheric haze, natural wind movement, professional camera quality. The setting is a beautiful hot-air-balloon launch site and aerial valley landscape with vineyards, winding rivers, mountains, villages, baskets, burner flames, and many colorful balloons. Main character: Use one female protagonist, {argument name="character name" default="the same young woman throughout"}, with shoulder-length dark brown hair, natural makeup, warm medium skin tone, soft oval face, and a friendly smile. She wears the same {argument name="wardrobe" default="ivory cream dress with a light beige cardigan"} in every image. Preserve her identity consistently in all scenes; no duplicate characters and no face drift. Layout: Each of the 6 scene cards should contain cinematic thumbnail frames along the top and small production notes underneath. Use black uppercase section labels and compact readable text. The thumbnails should alternate between drone-wide landscape views, close-up portrait/selfie shots, basket/burner details, and aerial balloon views. Scene cards: Include exactly 6 numbered storyboard sections: 1. Sunrise Introduction (0–4 sec): three thumbnails showing a sunrise drone-wide view of dozens of balloons, the woman walking toward the launch area, and a close selfie-style smiling portrait with balloons behind her. Notes: Scene/action: drone wide reveal of balloons at sunrise; cut to walking selfie as she approaches the launch field, smiling. Dialogue: {argument name="opening dialogue" default="Good morning! Today we're finally going on a hot air balloon adventure!"} 2. Preparing for Takeoff (4–8 sec): four thumbnails showing colorful balloon fabric inflating, realistic burner flames, the woman watching excitedly beside the balloon, and her near/inside the basket. Notes: crew inflates balloons, burner fires, she talks to pilot, watches balloon expand, and steps into the basket. Dialogue: “I can't believe how massive these balloons are up close!” Camera/movement: wide, close-up burner, gimbal follow, POV entering basket. 3. Lift-Off (8–13 sec): three thumbnails showing a close-up portrait of the woman, a side profile at the basket rail, and a wide aerial landscape with balloons floating. Notes: balloon lifts off, she records selfie and looks over edge as the landscape shrinks below. Dialogue: “We're actually floating... this is unbelievable!” Camera/movement: selfie, drone orbit, over-shoulder, wide aerial shot. 4. Floating Above the Valley (13–18 sec): three thumbnails showing a high aerial valley with river and balloons, a close portrait in the basket, and a POV view over vineyards and villages. Notes: golden sunlight over valley, rivers, vineyards, and hills; she admires the view and points at distant balloons. Dialogue: “This has to be one of the most beautiful views I've ever seen.” Camera/movement: 360° orbit, close-up, ultra-wide drone, POV over basket edge. 5. Magical Mid-Flight Moments (18–24 sec): three thumbnails showing basket ropes/burner hardware in close-up, the woman holding the camera selfie-style, and her waving toward another colorful balloon nearby. Notes: she takes photos, waves to other ballooners, and enjoys peaceful mid-flight moments. Dialogue: “I never imagined it would feel this peaceful.” Camera/movement: macro, handheld selfie, wide aerial, tracking side angle. 6. Landing & Farewell (24–30 sec): three thumbnails showing the balloon landing with crew silhouettes, a smiling farewell portrait, and a final aerial view of balloons in the sunrise haze. Notes: balloon lands, she steps out, thanks the crew, and waves goodbye as the drone pulls away. Dialogue: “What an unforgettable experience. Thanks for coming along, and I'll see you in the next adventure!” Camera/movement: wide landing, medium follow, close-up, drone pull-away. Bottom technical strip: Include exactly 5 labeled information boxes across the bottom plus a small centered character portrait near the identity box. The 5 boxes are: Camera & Cinematography, Audio, Environment, Character (Locked Identity), and Negative Prompt. Fill them with compact bullet-style text: Sony FX3, RED V-Raptor, DJI Ronin 4D, DJI Mini 4 Pro drone, 24mm/35mm/50mm/65mm lenses, handheld/gimbal/drone/POV/orbit, 4K HDR, shallow DOF, natural light; natural dialogue, wind, burner flames, birds, crew conversations, basket creaks, footsteps, soft cinematic orchestral music; sunrise hot-air balloon site, colorful balloons, wicker baskets, burners, valleys, rivers, vineyards, mountains, golden-hour light, realistic weather and atmosphere; use uploaded reference as single source, same face/hair/skin tone/makeup, ivory dress and beige cardigan, natural expressions only, no identity drift or wardrobe changes; no face distortion, no identity drift, no duplicate characters, no broken anatomy, no extra fingers, no deformed hands, no flickering, no frame warping, no inconsistent wardrobe, no logos, no watermarks, no subtitles, no timestamps, no UI overlays. Constraints: The final image must read as a polished production storyboard, not a movie poster. Keep all 6 scene cards visible and numbered, keep the bottom 5 technical boxes, use consistent typography, preserve the same woman and outfit in every thumbnail, and avoid adding extra scenes or extra characters as protagonists.

Goal: Create a polished 3D scientific educational infographic titled {argument name="headline text" default="DNA Nucleotides"}, showing a close-up view of a DNA double helix with labeled nucleotide bases and molecular structure. Canvas: Landscape 4:3 composition, bright white laboratory-style background with soft depth-of-field molecular clusters in pale blue, lavender, mint, and purple around the edges. Use a clean, high-resolution textbook/medical visualization style with realistic glossy translucent spheres and rods, soft shadows, and shallow depth of field. Layout: Center the title at the top in large dark navy bold type, with the subtitle {argument name="subtitle text" default="A close-up view of the DNA double helix"} directly beneath in smaller medium-blue type. Fill the frame with a large twisting DNA double helix running from the left foreground through the center to the right side, with a second orange strand crossing diagonally through the lower right foreground. Keep the central base-pair region open and readable for labels. Subject details: Render the DNA as a molecular ball-and-stick model combined with translucent bead-like backbone surfaces. The sugar-phosphate backbone should be two twisting chains made of semi-transparent bluish-gray spheres with orange/gold connector rods and small red atoms. In the center, show exactly 4 labeled nucleotide bases: Adenine (A) in green at upper left center, Thymine (T) in purple at upper right center, Guanine (G) in blue at lower left center, and Cytosine (C) in orange at lower right center. Adenine pairs with Thymine across the top using dotted pale blue hydrogen bonds, and Guanine pairs with Cytosine below using three dotted pale blue hydrogen-bond rows. Molecular rings should be flat aromatic ring structures with small colored atom spheres and white hydrogen atoms. Text content and labels: Include exactly 7 callout labels with thin black pointer lines: “Adenine (A)” in green pointing to the green upper-left base; “Thymine (T)” in purple pointing to the purple upper-right base; “Guanine (G)” in blue pointing to the blue lower-left base; “Cytosine (C)” in orange pointing to the orange lower-right base; “Sugar-phosphate backbone” in black pointing to the right-side helical backbone; “Hydrogen bonds” in black pointing to the dotted bond lines between base pairs; and “Emerging RNA” in orange pointing to the orange strand crossing the lower foreground. Use crisp sans-serif typography and avoid crowding. Visual style: Hyper-detailed scientific 3D render, glossy translucent molecular beads, realistic refraction, soft bokeh background, professional biology textbook infographic, balanced educational clarity and cinematic depth. Constraints: Use exactly the 4 nucleotide labels and exactly the 3 structural labels listed above; do not add extra labels, icons, captions, legends, watermarks, or borders. Keep all text sharp and correctly spelled.

Goal: Create a wide 16:9 Japanese isometric infographic explaining an AI-agent ecosystem that completes landing page production through a “command tower” model, with the headline {argument name="headline text" default="AIエージェント：LP制作を完遂する「司令塔」のエコシステム"} across the top. Canvas: Horizontal slide, light sky-blue background with soft clouds, simple city-map roads, tiny trees, pedestrians, cars, and technology buildings. Use a clean blue, cyan, white, and black palette with thick rounded vector outlines, friendly SaaS/tech illustration style, crisp readable labels, and subtle halftone texture. No photorealism. Layout: Center a large circular roundabout containing the main AI Agent command-center building, with a domed roof, clock, gears, and brain icon. Around it place exactly 4 main workflow buildings/cards connected by roads and directional arrows: 1) PLAN at upper left, 2) RESEARCH at upper right, 3) CREATE at lower left, 4) TEST & APPROVE at lower right. Add a final publication destination on the far right: a storefront labeled as the publishing/operation start point with a rocket launch building nearby. Include supporting city objects: exactly 3 satellite dishes on the research building roof area, exactly 3 database/server stack icons near the upper right, exactly 2 tall office towers at the far right, exactly 1 small bank/store building at far upper left, exactly 1 office block at bottom left, and exactly 1 foreground dome building at bottom center. Text content: Use Japanese text throughout. Top title must read {argument name="headline text" default="AIエージェント：LP制作を完遂する「司令塔」のエコシステム"}. Central circle label: “AI AGENT：理解・計画・判断を担う中心地”. Inside the central building, add a small panel saying the agent understands user goals and automatically assembles necessary tasks. Left side user note: “ユーザー依頼：LPを作って！” Sections: Show exactly 4 labeled production modules with these visible labels and contents. PLAN module: title “PLAN：依頼を5つのタスクへ分解”; list exactly 5 numbered items: 1 調査, 2 構成, 3 制作, 4 検証, 5 公開; add a caption saying the request is decomposed into five tasks and execution priority is determined. RESEARCH module: title “RESEARCH：外部と内部の情報を統合”; show exactly 2 content icons labeled “Webデータ” and “内部資料”; caption about integrating competitor, market, and internal product information. CREATE module: title “CREATE：複数ツールの使い分け”; show exactly 3 tool icons labeled “キャッチコピー AIライティング”, “画像生成 AIデザイン”, and “コーディング AI開発”; caption saying copy, images, and coding are produced using appropriate specialized tools. TEST & APPROVE module: title “TEST & APPROVE：品質検証と人間の承認”; show exactly 3 verification icons: an AI chip, a checked monitor, and people/approval; include labels for display check, quality check, and brand/person-in-charge approval; caption saying AI handles display and quality checks and the final human approval begins publication. Subject details: Roads should visually connect every module to the central AI Agent hub, emphasizing relationships rather than a strict linear flow. Place the publish destination on the right with an arrow leading to a striped storefront labeled “公開・運用開始”. Add small decorative tech details such as antennas, document icons, code brackets, magnifying glass, gears, cloud shapes, and tiny people walking between buildings, but keep them secondary. Style constraints: Isometric 2.5D diagram, flat vector illustration, blue monochrome emphasis, black outlines, clean infographic composition, high legibility, no watermark except a tiny unobtrusive “Gemini Notebook” style credit in the bottom right if needed. Keep all labels compact and aligned inside white rounded panels.

Goal: Create a clean Japanese promotional thumbnail/banner for a GPT Image 2 tutorial about using reference images to keep the same person consistent. Canvas: 16:9 horizontal banner, 1200×675 style, soft pale blue and white background with large organic blob shapes. Use a polished educational course-thumbnail aesthetic, airy spacing, high readability, no harsh shadows. Layout: Split composition with a realistic portrait on the left occupying about 40% of the canvas, and a text-heavy information panel on the right occupying about 60%. The right side has a large white rounded organic shape as the main text area. Add small decorative dotted grids in the top-right and bottom-left corners, a light diagonal hatch pattern near the lower-right edge, and coral accent marks beside the headline. Subject details: On the left, show one smiling Japanese woman in her late 20s to early 30s, photographed from the waist/chest up, facing the camera. She has shoulder-length softly layered black hair, natural makeup, warm friendly expression, and wears a peach sheer button-up shirt over a white ribbed camisole with a delicate gold necklace. Background behind her is bright, soft-focus, and slightly green-white like daylight through a window. Text content: Include exactly 4 text blocks on the right: 1) a dark navy rounded pill at the top reading 「GPT Image 2 実践編 ③」; 2) a large navy Japanese headline reading 「基準画像を\n作った、その次は？」 with a coral underline under the final phrase; 3) a thin outlined rectangle subtitle reading 「人物固定の具体的な使い方」 with 「具体的な」 highlighted in coral; 4) a bottom checklist row with a pale blue circular check icon followed by 「毎回どの基準画像を使う？」. Keep all Japanese text crisp, centered within its area, and use elegant Mincho-style serif for the main headline with modern sans-serif for smaller labels. Color palette: Deep navy text, pale powder blue background, white panels, peach/coral accents matching the woman's shirt, and light blue-gray icon details. Constraints: Use exactly one person, exactly four text blocks, exactly one check icon, and exactly three main decorative accent types: dotted grids, coral rays/underline, and diagonal hatching. Avoid extra logos, watermarks, English translations, or additional text.

Blueprint for a robot from the '{argument name="organization name" default="Metropolitan Police Heavy Equipment Task Force Robot Unit"}'

Using REFERENCE_0 as the base image, keep the castle artwork unchanged and overlay rough handwritten edit notes directly on top of it, as if marking areas for an image-editing revision. Add exactly 2 annotation groups: 1. Red annotation group on the tall rear tower near the upper-right center: draw red diagonal hatching across the tower, add a red arrow pointing toward it, and add the handwritten Japanese note {argument name="red edit note" default="一番奥の高いこの塔を崩して、無くして空にして。"} in the upper-right sky area. 2. Green annotation group around the crumbling edge/remaining structure near the middle top: draw a green irregular outline around the broken area, add a green arrow pointing down to it, and add the handwritten Japanese note {argument name="green edit note" default="この辺崩れのこっている"} in the upper-middle sky area. Style constraints: the annotations should look like quick digital pen markup, not integrated into the painting; keep the original watercolor fantasy castle image otherwise intact with no actual architectural removal yet.

Goal: Create a clean comparison graphic showing three AI-generated urban billboard photos side by side, as if comparing image model outputs. Canvas: Wide 16:9 presentation slide with a warm off-white background and generous margins. Place three tall vertical photo panels evenly spaced across the canvas, each framed by a pale beige mat/border. Under each panel, add a centered caption. Layout: Exactly 3 vertical street-photo panels in one row. Panel 1 on the left, Panel 2 in the center, Panel 3 on the right. Each panel shows a realistic New York City street scene with parked cars in the foreground, tall buildings, sidewalks, traffic signs, and a large square billboard mounted high on a building. Use slightly different street angles and lighting in each panel, like three separate model outputs of the same concept. Panel details: Left panel: a rear view of a black SUV parked on a city street, looking toward storefronts and buildings; a large white billboard above the street reads like a social-media quote card with a small circular animal avatar, the heading “What do you think?”, a short quote, and small ChatGPT branding. Center panel: a black SUV parked on a narrow Manhattan street with the Empire State Building visible in the distance; a large dark billboard with white text, a circular animal avatar, and ChatGPT branding. Right panel: a black sedan parked along a busier avenue with storefront signs such as “I ♥ NY GIFTS”; a large black billboard with a bright cyan border, white quote text, a circular animal avatar, and ChatGPT branding. Caption text: Use exactly 3 captions beneath the panels. Left caption: “ChatGPT Images 2.0” on the first line and “MEDIUM” in bold quotation marks on the second line. Center caption: “mona-lisa-1” in bold quotation marks. Right caption: “ChatGPT Images 2.0” on the first line and “HIGH” in bold quotation marks on the second line. Customizable text slots: The comparison labels should read {argument name="left model label" default="ChatGPT Images 2.0"}, {argument name="center model label" default="mona-lisa-1"}, and {argument name="right model label" default="ChatGPT Images 2.0"}. The left quality tag should be {argument name="left quality tag" default="MEDIUM"} and the right quality tag should be {argument name="right quality tag" default="HIGH"}. Visual style: Photorealistic smartphone street photography inside each panel, but arranged as a polished social-media comparison slide. Keep the overall slide minimal, with no decorative graphics beyond the three framed images and their captions. The billboard text inside the photos may be small and partially imperfect, but it should clearly resemble AI quote-card advertising with ChatGPT logos.

Goal: Create a dark gothic hand-painted infographic poster titled {argument name="headline text" default="THE AI STACK"}, showing an interconnected personal AI-tool stack as a mysterious occult network. Canvas: Square 1:1 composition, black and charcoal textured canvas with visible rough brush strokes, scratched chalk marks, muted cream ink, and subtle sepia highlights. The mood is eerie, nocturnal, vintage, and slightly haunted. Layout: Place a large central circular node labeled {argument name="main tool" default="CLAUDE"} with a simple eight-point starburst symbol inside it. Four rounded rectangular tool cards connect to the central node with pale hand-drawn cables and small circular connector dots. Use exactly 4 outer cards: top-left card labeled {argument name="tool 1" default="CODEX"} with a browser-window code icon, top-right card labeled {argument name="tool 2" default="GPT IMAGE 2"} with a framed landscape/image icon, bottom-left card labeled {argument name="tool 3" default="ANTIGRAVITY"} with a starry mountain or sparkling night icon, and bottom-right card labeled {argument name="tool 4" default="HERMES"} with a crossed-arrows or janitor-like routing symbol. Keep all text in uppercase, widely spaced, distressed serif/typewriter lettering. Scene details: Add a glowing full moon centered above the main node, a small spiderweb and tiny spider in the upper-left corner, a dark gothic castle silhouette in the right background, fog and smoky clouds along the bottom, a wilted flower with a bent stem leaning across the left side of the central node, one fallen brown leaf near the base of the flower, and an antique skeleton key lying at the bottom right. Include a tiny creator handle centered at the bottom: @AunySillyMe. Visual style: Painterly gothic chalkboard illustration, rough impasto texture, off-white linework on black, muted gold-brown accents on the dead flower and leaf, softly glowing moon, handmade imperfect borders, no clean vector polish. Constraints: Use exactly 1 central circle, exactly 4 connected outer cards, exactly 4 outer tool labels, exactly 1 moon, exactly 1 spiderweb, exactly 1 spider, exactly 1 wilted flower, exactly 1 fallen leaf, exactly 1 castle silhouette, and exactly 1 key. Do not add extra cards, extra tools, bright colors, modern UI gloss, or photorealism.

Goal: Create a wide comparison image showing exactly 4 separate Japanese infographic poster pages side by side, each explaining {argument name="topic" default="Cloudflare WebMCP"}. The posters should look like polished AI-generated tech explainer slides for social media, with dense but readable Japanese copy, Cloudflare branding, icons, diagrams, and section cards. Canvas: Extra-wide horizontal canvas, approximately 3:1 aspect ratio, with 4 tall white infographic pages arranged left-to-right. Each page is separated by thin black vertical gaps or borders. Each page has a clean white background, rounded pastel section boxes, colored headers, small icons, arrows, and a bottom-right semi-transparent dark rounded button overlay on some pages. Overall visual style: Modern Japanese SaaS/technical infographic design, Cloudflare orange-and-blue palette, flat vector icons, thin gray borders, rounded cards, tiny code blocks, flow diagrams, checkmarks, warning triangles, shield/security icons, robots, browser windows, AI agent icons, databases, clouds, and network arrows. Use a crisp printed-slide look with very small but believable Japanese text. Include Cloudflare cloud logo marks and the word “cloudflare” near the top of each page. Highlight “WebMCP” in orange. Page layout: Include exactly 4 infographic pages: 1. Left page: Title “Cloudflare WebMCPとは?” with orange “WebMCP”, subtitle about WebMCP as a bridge connecting web sites and AI agents. Include exactly 6 numbered horizontal sections: 1) “WebMCPって何?” with a simple web page to AI-agent diagram and green “ポイント” bullets; 2) “なぜ必要なの?” with red X problem bullets and orange solution bullets; 3) “WebMCPの仕組み” with a 5-step icon flow; 4) “Webサイト側の実装イメージ” with a dark code snippet block on the left and benefit cards on the right; 5) “セキュリティ” with green checklist and Chrome/WebMCP tool card; 6) “まとめ: WebMCPのメリット” with four benefit icons. 2. Second page: Title “Cloudflare WebMCPとは?” with a small orange Cloudflare logo and explanatory paragraph. Include exactly 5 numbered sections: 1) “WebMCPとは?” showing AI agent, web application, and Cloudflare WebMCP connection; 2) “仕組み: WebMCPはどのように動作する?” with a left-to-right flow from AI agent to WebMCP endpoint to user approval to tool execution; 3) “Cloudflare が WebMCP を簡単にできる理由” with purple feature cards and a code snippet; 4) “導入の流れ” with orange step cards for Workers/Pages, API definition, endpoint, MCP setup, and AI use; 5) “セキュリティと権限の考え方” with red warning/security cards. Add a dark pill overlay at lower right reading “もっと生成”. 3. Third page: Title “Cloudflare WebMCPとは?” with subtitle about publishing tools on web sites for AI agents. Include exactly 7 numbered sections with compact cards: 1) “WebMCPの主な概念” with three cards for AI agent, WebMCP, and security/permission; 2) “Cloudflareが提供するWebMCPの仕組み” with a detailed architecture diagram using browser, Cloudflare, worker, API, and data-store icons; 3) “WebMCPの活用例” with multiple small use-case cards; 4) “なぜCloudflareでWebMCPが使いやすい?” in an orange band with benefit icons; 5) “実装例: シンプルなサイト連携” with a small form/table and tool cards; 6) “導入ステップ” with a 5-step horizontal flow; 7) “セキュリティのポイント” with green checklist cards. Add a dark pill overlay near the bottom reading “さらに深く考える”. 4. Right page: Title “Cloudflare WebMCPとは?” with an orange Cloudflare logo and a small browser/address-bar illustration at top right plus an AI agent icon. Include exactly 6 numbered sections: 1) “WebMCPとは?” with definition bullets and a diagram from web app to AI agent; 2) “なぜCloudflareがWebMCPを推進している?” with exactly 4 cards describing AI-friendly web, security, open standards, and Cloudflare ecosystem; 3) “WebMCPの仕組み” with exactly 5 icon steps from site/API to MCP endpoint to AI request to tool execution to result; 4) “WebMCP manifest のイメージ” with a dark code block and a green checklist card; 5) “セキュリティとして重要” with warning and permission cards; 6) “まとめ: WebMCPの価値” with four blue/purple benefit cards. Text content: The main headline on every page must include {argument name="headline text" default="Cloudflare WebMCPとは?"}. Keep all visible body text in Japanese. Use credible Japanese labels such as “AIエージェント”, “Webサイト”, “MCPサーバー”, “ツール定義”, “認可”, “権限管理”, “Workers”, “Pages”, “API”, “manifest”, “セキュリティ”, and “導入ステップ”. Use tiny text where necessary but keep major section titles legible. Branding and colors: Use {argument name="brand name" default="Cloudflare"} branding, orange cloud logos, blue section headers, green success boxes, orange highlight boxes, purple feature boxes, red security/warning boxes, and gray explanatory cards. Maintain a professional Japanese tech-blog infographic aesthetic. Constraints: Show exactly 4 pages, exactly 6 sections on page 1, 5 sections on page 2, 7 sections on page 3, and 6 sections on page 4. Do not include people or photorealistic scenes. Do not add watermarks. Make the image information-dense, like a generated social-media comparison of four alternate infographic designs.

Create a cinematic photorealistic sci‑fi scene inside a pristine futuristic laboratory, showing a glossy white humanoid android seated at an advanced command workstation, viewed from a three-quarter side angle. The android has a smooth featureless black glass face visor, rounded helmet-like head, sleek white synthetic body panels, subtle glowing cyan seams along the arms, torso, legs, and head, and the word {argument name="android chest label" default="SINGULAR"} faintly illuminated on its chest. It sits in a modern ergonomic white chair with gray armrests, both hands hovering over and touching a glowing transparent touchscreen console. The lab is bright white and silver with curved wall panels, soft blue-white LED strips, clean reflective floors, and depth-of-field background monitors. The workstation should contain exactly 7 major display areas: 1 large left vertical monitor labeled {argument name="left monitor title" default="QUANTA NETWORK"} with a blue node-and-edge quantum network diagram and an entanglement rate chart below it; 1 central upper monitor labeled {argument name="main lab title" default="QUANTUM MATERIALS LAB"} with system overview charts, circular status gauges, line graphs, and system metrics; 1 right upper monitor panel with multiple cards for crystal phase space, AI systems diagnostics, and material structure; 1 wide horizontal transparent main console under the robot’s hands with glowing blue interface panels, graphs, and a central network visualization; 1 small floating holographic display above the console showing dense technical data and a miniature chart; 1 angled auxiliary touchscreen on the far left foreground with a circular blue scanner interface; and 1 small embedded control tablet near the back of the desk with colorful buttons. Use readable futuristic UI text sparingly, including labels such as {argument name="diagnostic label" default="AI SYSTEMS DIAGNOSTICS"}, {argument name="materials label" default="MATERIAL STRUCTURE"}, “SYSTEM OVERVIEW,” “SYSTEM METRICS,” “QUANTUM LINK PROTOCOL,” and “ENTANGLEMENT RATE.” Visual style: ultra-detailed 3D render, photorealistic hard-surface design, glossy white polymer and brushed metal materials, transparent glass UI surfaces, luminous cyan and electric blue highlights, small orange/yellow chart accents, shallow depth of field, volumetric soft lighting, high-tech medical-clean atmosphere. Composition should emphasize the android on the right and the dense quantum-computing dashboard wall on the left, with realistic reflections and no clutter. Avoid extra characters, logos, watermarks, distorted hands, or unreadable oversized text.

Create an abstract topographic mixed-media artwork on a horizontal {argument name="aspect ratio" default="4:3 landscape"} canvas, resembling a large aerial geological map or energetic convergence diagram. The composition centers on a bright glowing circular hub slightly below the middle, made of concentric translucent rings in warm gold, cream, and gray, with thin radial lines extending outward. From this center, four broad organic terrain-like regions flow outward: a dark charcoal and slate-black river-shaped mass sweeping from the left edge into the center and down toward the lower left, a rusty burnt-orange and sienna canyon-like band expanding from the center to the lower right, a pale cream and sand-colored open field filling the upper and right areas, and muted gray-beige layered landforms in the upper left and lower middle. Cover the image with delicate contour lines, hairline curves, dotted paths, small node circles, and fine white and black route-like lines that weave across the colored regions. Include exactly 3 small orange dots near the upper center, exactly 2 large ochre circular stains in the upper right, and numerous tiny scattered black, white, gray, and gold dots arranged like constellations and map markers. Use a restrained palette of {argument name="main palette" default="ivory, charcoal black, warm gray, ochre gold, burnt orange, rust brown"}, with watercolor washes, mineral textures, translucent paper layers, ink scratches, stippling, and subtle grain. The style should feel like an elegant abstract cartographic painting, a cosmic topography, or a geological energy map, with no readable text, no labels, no border, no people, and no recognizable objects. Lighting should be soft and diffuse, with the central hub as the brightest point, creating a sense of motion and convergence. Render highly detailed fine linework and tactile handmade textures, suitable for a gallery-quality generative art print.

Build a modern flat graphic around {argument name="subject" default="any chosen subject object"}, with a huge Chinese theme character, geometric color blocks, and a single-line organic contour interlacing through one another. Keep an overall bright, clean background field, letting approximately two-thirds or more of the space remain empty, and concentrate the main weight in an off-center cluster: a saturated rectangular block in the theme's primary color first establishes a stable plane; a deeper, low-lightness, cool-toned circle overlaps from beneath it and extends beyond the boundary; the core Chinese theme character advances vertically in oversized, heavy, geometric black strokes across the white background, rectangle, and circle, with partial occlusion and cropping creating a strong scale jump; on the left, the date and a bilingual subtitle form a compact numeric column, while on the right, short vertical sentences set in two stroke-weight tiers balance the weight; a few miniature captions, short dashes, and linear symbols scatter along the edges, creating a three-level reading rhythm from main character to side notes to details. Abstract the dynamic relationship derived from the theme into one continuous organic contour, drawn as a broad sweeping loop with a clear, near-black single line that enters from the large empty field, passes through the color blocks, and crosses the edge; the curve intersects itself several times while remaining legible as a single path; keep most of its interior as the underlying background color, adding only localized, transparent, fading fills in the theme's primary color at the tips and extended sections, so the light line creates tension with the heavy giant character. Adapt the color scheme to the theme: let a clean, high-lightness neutral background carry the largest area; keep the theme's primary color vivid and highly saturated, using it for the rectangle and the localized transparent gradient; use a deep cool support color for the circle, separated in hue from the primary color and blending softly into the primary color at its end; use near-black for text and the contour, with a very small amount of light text reversed onto the primary color block; fix unequal areas, lightness hierarchy, and saturation differences, but do not fix specific color names. Render all elements as flat surfaces like matte digital printing, with sharp contour edges and gradients only as two localized, slowly feathered transitions; do not add dimensional shadows, paper grain, or global lighting effects. Keep typography in a low-contrast, broad, thick, near-geometric black sans-serif skeleton; allow the main character's parts to be elongated, offset, cut out, and switched between color-block backgrounds; give numbers a rounded futuristic structure and compress small text with narrow letterforms and loose tracking. Replace brand names, advertising copy, and concrete dates with new theme information. Preserve in the final result the extreme white

Goal: Create a cinematic dark tech launch banner for {argument name="product name" default="Grok Imagine Image 2.0"}, showing an advanced AI image editing workflow on a large futuristic display. Canvas: Wide 16:9 horizontal composition, glossy black studio environment, high contrast, premium product-announcement style. Use a deep charcoal background with subtle blue and magenta rim lighting, reflections on a polished desk surface, and a soft neon glow around interface elements. Layout: Place a large bold white headline centered at the top reading “{argument name="product name" default="Grok Imagine Image 2.0"}”. Below it, create one continuous widescreen glass UI display divided into exactly 3 vertical sections by thin glowing separators: cyan separator between the left and center sections, magenta separator between the center and right sections. Left section: Show local image editing. Include exactly 3 stacked reference thumbnails on the far left: 1) sunlit mountain peak, 2) dramatic clouds, 3) gray rock texture. To the right of these thumbnails, show a large mountain image with a dashed rectangular selection box, white corner/edge handles, and a glowing blue circular edit target over the mountain. Add a cyan arrow pointing from the cloud thumbnail area toward the mountain, implying guided editing. Center section: Show multi-image reference generation. Use a large realistic image of a modern glass house on a rocky coastal landscape at dusk, with a pool reflection in the foreground. Above it, include exactly 3 small rounded reference cards: 1) modern house exterior, 2) lone windswept tree on rocks, 3) pink-purple sunset sky over water. Connect the first card with a cyan dashed curved line ending in a glowing cyan dot, connect the second card with a vertical dotted line, and connect the third card with a magenta dashed curved line ending in a glowing magenta dot. Right section: Show intelligent resizing or outpainting. Include a sunset seascape image on the left side of the panel, a vertical boundary line at its right edge, and a larger dashed rectangular extension area to the right with a faint grid overlay. Add a white arrow pointing right across the extension area to indicate expansion. The right side should fade into a magenta glow. Foreground objects: On the desk below the display, include exactly 4 objects: 1) a small black cylindrical control puck with a cyan light ring on the left, 2) a slim dark drawing tablet centered with a soft blue rim glow, 3) a black stylus lying diagonally on the tablet, 4) two matte geometric props on the right consisting of one cube and one sphere. Visual style: Ultra-polished 3D render mixed with realistic UI mockup, dramatic cinematic lighting, glass panels, crisp rounded thumbnails, subtle grid textures, neon cyan and magenta accents, shallow reflections, high-resolution product keynote aesthetic. Keep the typography clean, bold, and sans-serif. No extra logos, watermarks, people, or additional text.

Goal: Create a vertical Chinese product comparison infographic on a white background with a clean flat-design style. The main title at the top is {argument name="headline text" default="2026 家庭算力采购指南"}, in large bold black Chinese type, centered. Canvas: Portrait 2:3 aspect ratio, approximately 768×1152 px. Use generous white margins, crisp vector lines, and a balanced grid layout. Layout: Place exactly 4 tall rounded-rectangle comparison cards in one horizontal row beneath the title. Each card has a colored header bar, a matching thin border, a simple line icon near the top, then five stacked information sections separated by pale gray horizontal divider lines. Use neat left-aligned Chinese text inside each card. Cards, colors, icons, and text content: 1. Blue card header: “Mac mini M4”. Icon: simple blue outline Mac mini with Apple logo and two small front dots. Sections: “型号:” / “M4 10核 CPU / 10核 GPU”; “参考价 (¥):” / “4,499 起”; “功耗 (W):” / “约 65”; “适合人群:” / “开发、轻量 AI、家庭服务器”; “一句话缺点:” / “内存与硬盘升级成本高”. 2. Green card header: “RTX 5090 主机”. Icon: green outline desktop tower with four fan circles and side panel. Sections: “型号:” / “RTX 5090 32GB”; “参考价 (¥):” / “30,000 起”; “功耗 (W):” / “约 800–1,000”; “适合人群:” / “本地大模型、渲染、重度创作”; “一句话缺点:” / “价格高，耗电与散热压力大”. 3. Purple card header: “云 GPU 包月”. Icon: purple outline cloud containing a server rack. Sections: “型号:” / “按需选择 GPU”; “参考价 (¥):” / “1,000–8,000 / 月”; “功耗 (W):” / “本地几乎为 0”; “适合人群:” / “短期项目、弹性需求、团队协作”; “一句话缺点:” / “长期使用成本不可控”. 4. Orange card header: “二手矿卡”. Icon: orange outline dual-fan graphics card. Sections: “型号:” / “RTX 3090 等”; “参考价 (¥):” / “3,500–6,000”; “功耗 (W):” / “约 350–500”; “适合人群:” / “预算有限、愿意折腾的玩家”; “一句话缺点:” / “寿命、保修与稳定性风险高”. Footer: Center a small gray footnote along the bottom: “数据仅供参考 · 更新于 2026-08-08 · 非投资建议”. Visual style: Minimal flat infographic, Chinese sans-serif font, high readability, accurate alignment, consistent spacing, no shadows, no photos, no decorative background. Ensure all Chinese text is sharp and legible, with the section labels colored to match each card’s header.

Goal: Create a single aged martial-arts manual page about {argument name="martial art title" default="降龍十八掌"}, designed like an antique Chinese kung fu secret manuscript with traditional Chinese typography, ink drawings, and worn parchment. Canvas: Square 1:1 page, full frame, yellowed parchment paper with dark burned edges, stains, foxing, creases, a vertical fold near the right side, thin double-line border, and black ink printing. Use a vintage woodblock/hand-inked style, slightly imperfect and weathered. Layout: Large bold brush-calligraphy title across the upper left reading {argument name="martial art title" default="降龍十八掌"}. Beneath it, a centered decorative subtitle line with small diamond ornaments reading {argument name="subtitle" default="總綱・心法・動作要領"}. At the upper right, place a vertical red seal stamp reading {argument name="seal text" default="本門秘傳"}, with a short vertical warning paragraph beside it. The page is divided into a left column of text instructions and a right column of action diagrams. Left column text: Include a short introductory paragraph in traditional Chinese. Below it, create a section titled 「心法要訣:」 with exactly 8 numbered black-circle items, numbered 1 through 8, each a one-line principle about breathing, posture, palm force, continuous motion, calmness, and dragon-like power. At the bottom left, add a boxed note section titled 「習練須知」 with exactly 3 bullet points about beginners practicing slowly, daily repetition, and avoiding misuse. Right column diagrams: Create a framed section titled 「動作要領圖示」 containing exactly 6 small instructional panels arranged in a 2-column by 3-row grid. Each panel shows the same robed martial artist in black ink demonstrating a different palm-strike stance, with curved motion arrows and small directional marks. Label the six panels in traditional Chinese as: 1. 「第一掌 龍騰勢」, 2. 「第二掌 潛龍勢」, 3. 「第三掌 飛龍勢」, 4. 「第四掌 神龍擺尾」, 5. 「第五掌 龍戰於野」, 6. 「第六掌 龍吟九天」. The figure wears ancient Chinese training robes, a tied belt, hair in a topknot, and dynamic wide stances. Bottom details: Along the bottom center, draw a small monochrome ink landscape of distant mountains, rocks, and bamboo. Add tiny footer text at the lower left and lower right in traditional Chinese, resembling chapter/page and secrecy notices. Visual style: Authentic antique Chinese manual page, high-contrast black brush ink on tan parchment, traditional Chinese characters, hand-drawn action illustrations, old print texture, subtle smudges, no modern elements, no color except black ink and the red seal. Keep the text dense but legible, with believable traditional Chinese manuscript formatting.

Goal: Create a cinematic sci-fi product mockup of a futuristic operations dashboard titled {argument name="product title" default="HOLOGRAPHIC WORKBENCH"}, floating as a transparent neon-blue holographic screen above a dark control desk in a cyberpunk command center. Canvas: Wide 16:9 landscape interface screenshot, dark navy/black glassmorphism UI, glowing cyan borders, subtle scanlines, tiny grid texture, high detail, realistic GPT-image-2 style render, shallow depth of field with blurred server lights in the background. The interface should feel like a clickable prototype built on top of a beautiful concept render. Header: Top-left brand text reads “PARACORTEX TERMINAL” above the main title “HOLOGRAPHIC WORKBENCH” with the subtitle “Situational awareness across code, agents, and infrastructure.” Top-center metadata shows exactly 3 fields: Workspace “paracortex-core”, Session “pipin@paracortex.dev”, Time “09:42:17 UTC”. Top-right includes a small button labeled “SIMULATE EVENT” plus tiny gear, sync, and plus icons. Left navigation: A vertical glowing sidebar with exactly 8 icon navigation items labeled, from top to bottom: Overview, Pulse, Agents, Jobs, Traces, Infra, Config, Reports. At the lower left, include a small system status panel with Status “Operational”, Health “98.6%”, and Uptime “170 4H 23M”. Top metric row: Show exactly 6 stat cards across the top. Card 1: “PULSE SCORE” value “78” status “GOOD” with a small line graph. Card 2: “ACTIVE AGENTS” value “42 / 64” status “ONLINE”. Card 3: “RUNNING JOBS” value “124” status “+12.4%”. Card 4: “SYSTEM HEALTH” value “98.6%” status “NOMINAL”. Card 5: “EVENTS (24H)” value “1,284” status “+8.7%”. Card 6: “TRACES (24H)” value “4.7M” status “+23.1%”. Main layout: Beneath the metric row, create exactly 5 large vertical columns/cards. Column 1 is “NEEDS YOU” with badge count 3 and exactly 3 task cards: “Release blocked” with urgent red styling, “Schema change detected” with yellow approval styling, and “PR #402 open” with purple lifecycle styling. Column 2 is “IN MOTION” with badge count 4 and exactly 3 visible task cards: “Indexer lagging” as a red incident with small x20 badge and tiny bar chart, “Refactor batch running” with teal progress bar, and “Reconnected” with blue session recovered styling. Column 3 is “CHANGED” with badge count 6 and exactly 3 task cards: “v1.4.0 shipped”, “Indexer catching up” with mini bar chart, and “Agent config updated”. Column 4 is “CONDITIONS” with badge count 4 and exactly 3 task cards: “Telemetry not monitored”, “Message bus healthy”, and “Agent pool healthy”. Column 5 is a right-side “SELECTED CONTEXT” details panel focused on {argument name="selected incident" default="Indexer lagging"}, showing a red x20 badge, a red histogram, a vertical timeline with timestamps, and affected service chips. Selected context details: The right detail panel should include title “Indexer lagging”, subtext “Since 09:25 UTC (16m)” and “20 repositories behind”. Add a red bar chart labeled along the bottom with times around 09:00, 09:15, 09:30, 09:45. Timeline contains exactly 5 entries: “09:42 Lag threshold breached”, “09:37 Alert sent”, “09:36 Auto mitigation started”, “09:34 Partial recovery”, and “09:26 Incident opened”. Add a small button labeled “Investigate”. Affected services chips: “paracortex-core”, “port-1”, “port-indexer”, “port-agent”, “+15 more”. Bottom activity area: Add an “ACTIVITY TRACE” panel spanning the width under the main columns, labeled “Last 60 minutes”. Include a horizontal time-based trace with multicolored dots, small filter tabs exactly 6: “All”, “Events”, “Commits”, “Jobs”, “Agents”, “Alerts”. On the right of this panel, include a compact event list and a tiny legend with exactly 6 categories: Alerts, Changes, Jobs, Commits, Infra, Info. Foreground controls: At the bottom center, place a separate rounded glass status dock floating in front of the screen. It contains exactly 4 large counters matching the columns: “Needs you 3” in red, “In motion 4” in blue, “Changed 6” in teal, and “Conditions 4” in amber. In the center bottom of the dock, include a glowing purple-blue cube orb emblem. Color and style: Use neon cyan outlines, teal and electric blue glow, red for incidents, amber for warnings, purple for lifecycle/session items. UI panels are translucent dark glass with rounded corners. Text is small but mostly legible, technical, dense, and polished. Add subtle reflections on the desk, small physical knobs and screens at the bottom edge, and atmospheric blue light beams. Constraints: Keep all listed counts exact: 8 sidebar items, 6 top metric cards, 5 main columns, 15 total visible main task/detail cards if counting the right selected context panel, 6 filter tabs, 6 legend categories, and 4 dock counters. Do not add people. Do not use a flat white background. Avoid cartoon styling; make it look like a premium sci-fi SaaS dashboard prototype.

A {argument name="aspect ratio" default="2:1 horizontal"} image. Divide the whole into left and right, with the left being 25% of the total width and the right being 75%. The right side is a 3:2 area using the full height, containing a three-view diagram on a {argument name="background color" default="gray background"} arranged side-by-side in the order of front, side, and back. The three-view diagrams should all be {argument name="standing poses" default="standing poses"} where the full body is visible. However, in all three cuts, the top should be cut slightly below the jaw, showing only from the jaw down. Everything above the neck is completely out of frame, and the face, head, and hair are not shown at all. The left side is divided into two parts, top and bottom, with two squares of the same size. Both are close-ups of faces on a gray background. The top shows a serious expression facing the camera, and the bottom shows a face looking away. The three-view diagram on the right is configured for full-body costume verification.

Using [{argument name="reference image" default="Image 1 Reference"}] as the sole visual basis, generate a vertical 3:4 {argument name="teaching topic" default="street photography knowledge analysis chart"} (educational infographic poster). Goal: Use clear and interesting diagrams to explain "subject position, camera position, lens/distance, camera height, shooting angle, and composition points" for the purpose of teaching others how to recreate the shot. Overall style: Clean and high-end photography tutorial infographic + light scrapbook-style annotations; light off-white paper texture or light gray background; mainly black, white, and gray; automatically extract 1–2 colors from Image 1 as accents; sharp lines, high information density but not cluttered, unified alignment, and strong readability. Mandatory Rules: - Must preserve the following from Image 1: subject pose, relative position between subject and environment, main background structural lines, lighting direction, and spatial logic. - Do not change the pose or the scene; only provide an "analytical diagrammatic presentation." - Faces are not the focus: they can be lightly blurred/simplified, no realistic beautification, no emphasis on facial features. - Full Chinese; no English, no brand names, no watermarks, no UI interface elements. Fixed Layout (remains the same every time, does not change with Image 1): 1) Main Visual Area (upper half or large image on the left): Place the "Reference Scene Display" (lightly stylized/illustrated based on Image 1, but with recognizable scene and posture), overlaid with: rule-of-thirds grid, subject outline box, leading line prompts, vertical/horizontal line correction prompts; use numbered dots + hand-drawn arrows to mark key points (subject landing point, negative space direction, leg extension direction, background leading line source). 2) Diagram Area (two side-by-side images on the right or bottom half): - Top View: Use minimalist icons to draw wall/street/building boundaries (abstracted from Image 1), subject position (person icon), camera position (camera icon); must show shooting direction arrows, distance between subject and camera (dashed line + numerical range), recommended position points A/B (two alternative positions); camera orientation must match Image 1 (front/side/oblique, etc.). - Side View: Draw ground line, subject height示意, camera height示意, tilt angle sector; must show camera height range from ground (reasonably inferred from Image 1), lens tilt angle range (reasonably inferred from Image 1); low angle/eye level/high angle must be correct. 3) Parameter Card Area (three fixed rounded-corner cards at the bottom): - Card 1: Lens and Distance (give recommended equivalent focal length range + shooting distance range). - Card 2: Camera Height and Angle (give camera height range + tilt angle range). - Card 3: Pose and Body Points (summarize key points of the pose in Image 1 using short sentences). 4) Error Demonstration Area (fixed at the bottom right): 3 red-cross error prompts (automatically generated based on the most relevant issues in Image 1, short sentences), for example: don't use ultra-wide angle too close to the face, don't crop toes/fingers, don't use a camera position that's too high resulting in distorted proportions (adjust automatically based on Image 1's actual situation). Automatic Inference Requirements (to make it usable for any replacement of Image 1): - Automatically infer and write into the diagrams based on Image 1: camera orientation, camera height, shooting distance interval, equivalent focal length interval, subject landing point and cropping strategy, background leading line source. - All numerical values should be expressed as "ranges" (credible and reproducible), do not provide a single fixed number. Output Requirements: High definition, sharp lines, natural shadows, unified icons, neat layout, readable information; no garbled text, no logos, no watermarks, no excessive decoration. Negative Constraints: No watermarks, no logos, no English, no garbled text, no overly realistic faces, no exaggerated cartoon proportions, no limb deformities/extra fingers, no cluttered layouts, no low resolution, no heavy filters, no changes to the subject pose or scene relationship of Image 1.

Goal: Create a friendly square educational promo illustration introducing {argument name="character name" default="ChatGPT"} as a cute AI assistant robot made by {argument name="company name" default="OpenAI"}, with a warm desk scene and hand-drawn annotation style. Canvas: Square 1:1 composition, high-resolution, cozy modern office lighting, shallow depth of field. The scene is centered on a glossy white robot mascot sitting behind a wooden desk, facing the viewer and waving. Main subject: A small rounded humanoid robot with a white plastic body, black glossy face screen, glowing happy crescent eyes, rounded ear-like side panels, black joint details, and a teal OpenAI-style knot logo on its chest. The robot holds a white placard in its left hand and raises its right hand in a friendly wave with five black-and-white segmented fingers. Add a handwritten arrow above the robot reading “That’s me!” pointing to its head. Placard text: The placard should show a teal knot logo at the top, then large bold text “OpenAI”, with handwritten text above it reading “I was created by”. Below the logo and company name, add “An AI research and deployment company”, then a teal handwritten slogan: {argument name="slogan" default="Better AI. A brighter future."} Left annotation panel: Add large chalk-like handwritten text in the upper left: “Hi there!” and beneath it “I’m” followed by the large mint-green underlined word “ChatGPT”. Under this are exactly 3 icon-and-text facts: 1) a speech bubble icon with “A large language model (AI assistant)”; 2) a lightbulb icon with “I can answer questions, explain things, help you create, solve problems, and much more!”; 3) a heart icon with “I’m here to be helpful, honest, and friendly.” Right background board: On a dark navy vertical board in the background, place a pale OpenAI-style knot icon and the word “OpenAI”. Below it, add exactly 4 stacked handwritten lines: “Research”, “Safety”, “Useful AI”, “for everyone”, plus small decorative underline marks. Desk objects: Include exactly 4 stacked books on the left side of the desk with readable spine titles, from top to bottom: “BETTER QUESTIONS”, “CLEARER ANSWERS”, “A MORE HELPFUL TOMORROW”, and a fourth dark book with a small heart mark. Put a partially visible silver laptop in the lower left foreground with a small handwritten sticker-like note: “Curiosity builds a better world” and a heart. Put one white coffee mug on the right with handwritten text: {argument name="mug text" default="Same team. A brighter future."} and a small smiley face. Bottom caption: Along the lower edge of the wooden desk, add a thin hand-drawn line and small handwritten text: “OpenAI × ChatGPT = More possibilities for everyone” followed by a heart. Visual style: A polished 3D-rendered mascot scene combined with whiteboard/chalk marker doodles, soft bokeh background, warm natural light, crisp readable typography, teal accent color, navy and wood tones, friendly optimistic tech branding. Use hand-drawn white and mint text overlays as if written on glass in front of the scene. Constraints: Keep all visible text in English and legible. Use exactly 3 left-side fact icons, exactly 4 books, exactly 1 laptop, exactly 1 mug, and exactly 1 robot. Avoid photorealistic humans, avoid clutter, avoid extra logos beyond the OpenAI-style knot marks, and do not add watermarks.

Goal: Create a Japanese comparison infographic slide about how AI image generation changes when given only the prompt {argument name="prompt phrase" default="幻想的な画像"}, comparing August 2025 and August 2026 outputs. Canvas: 4:3 horizontal presentation slide, dark navy gradient background, polished tech-report style. Use crisp white Japanese typography, glowing borders, and subtle blue/purple highlights. Overall mood should feel like a serious visual analysis of fantasy AI art evolution. Header: At the top center, large bold Japanese title: 「『幻想的な画像』とだけ指示したときの比較」. Directly underneath, smaller subtitle: 「AI画像生成の進化：2025年8月 vs 2026年8月」. Layout: Use exactly 2 large rounded comparison panels side by side. - Left panel: blue theme, labeled 「2025年8月（GPT Image / 従来世代）」. - Right panel: purple theme, labeled 「2026年8月（ChatGPT Images 2.0）」. Each panel has a pale prompt strip near the top reading 「プロンプト：幻想的な画像」. Left panel content: Show exactly 1 fantasy artwork: a traditional painterly fantasy landscape with a huge full moon, deep blue night sky, forest silhouettes, a winding glowing river, small firefly-like lights, and a distant dark castle with spires. It should look like a complete single illustration, warm and nostalgic, with a restrained composition. Below the artwork, create exactly 2 analysis boxes: 1. Box title 「特徴」 with exactly 3 bullet points: 「月・森・城の定番モチーフに収束」, 「全体にやや黄ばんだ暖色の光」, 「シンプルで一枚の完成形として出力」. 2. Box title 「色調の傾向」 with a small horizontal color palette of exactly 4 swatches: deep blue, teal green, warm yellow, golden ocher. Add explanatory text: 「青＋緑＋黄のバランスで、古典的ファンタジー絵画のような温かみのある色調」. Right panel content: Show exactly 2 fantasy artworks side by side inside the panel, separated by a thin vertical divider. 1. Left artwork: a luminous fantasy city at night with a crescent moon, starry sky, distant towers, glowing waterfront, and a long-haired woman in a flowing dress standing on a balcony or bridge, looking toward the city. 2. Right artwork: an opulent celestial castle floating above clouds, white and lavender architecture, sparkling sky, crystalline towers, pink-purple blossoms, and a small robed figure looking into the glowing city. The right panel art should feel more complex, cinematic, saturated, and concept-art-like than the left panel. Below the artworks, create exactly 2 analysis boxes: 1. Box title 「特徴」 with exactly 3 bullet points: 「複数の解釈を提示（2パターン）」, 「天空都市・星空・人物などより多様で壮大な構図」, 「色彩が鮮やかで幻想性が高い」. 2. Box title 「色調の傾向」 with a small horizontal color palette of exactly 5 swatches: cyan blue, violet, pale pink, white, light sky blue. Add explanatory text: 「青＋紫＋ピンク＋白が中心で、デジタルファンタジー／コンセプトアート的な冷色系」. Bottom summary bar: Across the full width at the bottom, add a rounded dark blue box. On the left, a small label button reading 「まとめ」. To the right, summary text: 「2025年は『典型的な幻想風景を一枚』に収束していたのに対し、2026年は『幻想的』という概念から複数の解釈を生成し、色彩・構図・世界観の幅が大きく拡張している。」 Visual style: Clean Japanese infographic, presentation slide, thin neon blue/purple borders, subtle drop shadows, high readability, balanced margins, precise alignment. Use fantasy images that look like embedded generated examples rather than photographs. Constraints: Include exactly 3 fantasy artworks total, exactly 2 main comparison panels, exactly 4 analysis boxes, and exactly 2 color palettes. Keep all Japanese text legible and do not add extra labels, logos, watermarks, or irrelevant decorations.

Goal: Create a realistic screenshot-style vertical Chinese social media post about a fun feng shui city-direction personality test. Canvas: Tall smartphone screenshot, approximately 9:16, white app interface background, iPhone status bar at the top showing time 18:17, signal, 4G, and battery. The whole image should look like a Xiaohongshu/RED note screenshot rather than a clean poster. Top app header: A back arrow on the left, a small circular female avatar, the account name {argument name="account name" default="星河小屋"}, a rounded pink-outline follow button labeled 「关注」, and a share arrow icon on the right. Add a small gray carousel badge near the upper right of the post image reading 「1/5」. Main post image: Center a colorful simplified map of China on a white background. Use pastel province blocks with Chinese province labels such as 新疆, 青海, 西藏, 四川, 云南, 贵州, 广西, 广东, 福建, 浙江, 上海, 安徽, 山东, 河南, 北京, 内蒙古, 辽宁, 吉林, 黑龙江, 台湾. Overlay a black feng shui/bagua compass diagram: one large circle, multiple radial black lines, a yin-yang symbol at the center, and bold Chinese five-element/direction labels including 水, 木, 火, 土, 金 and direction notes such as 正北, 正南, 正东, 正西, 西北, 东北, 东南, 西南. The map should feel like a casual educational graphic, slightly imperfect but readable. Main headline text: At the top of the post image, use very large dark brown-black Chinese calligraphy-style title text: {argument name="headline text" default="哪个方位最旺你"}. Below it, centered smaller black text: {argument name="subtitle text" default="换个城市顺风顺水!"}. Post body under the image: Add Chinese caption text in a social feed layout. First line in larger black text: {argument name="caption hook" default="一定要去旺你的城市走一走！五道题测出"}. Then smaller lines: 「身体其实早已替你锁定专属城池」, 「你会不会有这样的瞬间：」, 「刷到一座城市的风景图，心头莫名悸动」, 「听见一座城市的名字，似曾相识格外熟悉」. Bottom engagement bar: At the bottom, include a rounded comment input field with a pencil icon and placeholder 「说点什么...」. To its right show exactly 3 engagement items: a heart icon with 「1452」, a star/bookmark icon with 「760」, and a comment bubble icon with 「1.4万」. Draw a thick hand-drawn red oval circle around these three engagement metrics, emphasizing them. Visual style: Realistic mobile screenshot, crisp UI icons, clean white margins, slight social-media compression, readable Chinese typography, map illustration in soft blue/green/yellow/orange/red provinces, black compass lines. Keep the image vertical and avoid adding extra panels or unrelated decorations.