Diagram / Chart

Goal: Create a wide editorial hero banner about Codex的下一块拼图, showing a jigsaw-puzzle metaphor for AI coding, visual design, and missing capabilities. Canvas: Ultra-wide horizontal banner, approximately 5:2 aspect ratio, dark charcoal technical background with faint grid lines, subtle blueprint marks, tiny dots, and thin circuit-like guide lines. Center the artwork with a large off-white paper/cardboard jigsaw puzzle assembly casting soft shadows. Layout: Use exactly 15 visible puzzle elements: 13 off-white connected puzzle pieces forming a rectangular mosaic, 1 vivid cobalt-blue puzzle piece inserted at the bottom-right edge, and 1 dashed-outline empty missing-piece slot immediately to the right of the blue piece. The puzzle should look physical, with rounded tabs and sockets, slight bevels, paper texture, seams, grime, and realistic shadowing. Text content: The main typography dominates the puzzle surface. Use huge bold black sans-serif Chinese type reading Codex的 across the top row and 下一块拼图 across the middle row. Make the letters oversized, cropped slightly by puzzle seams, and integrated into the tile layout. Keep the text crisp and high contrast. Visible graphic elements: Include exactly 7 small embedded UI/diagram motifs on puzzle pieces: 1 dark code editor panel in the lower-left with short colored code lines and a small </> symbol; 1 pie chart combined with a small bar chart along the lower middle; 1 simple node-and-connector flow diagram near the lower center-right; 1 bullet-list panel with three round bullets to the right of the flow diagram; 1 user profile card in the upper-right with avatar icon and horizontal text lines; 1 small spreadsheet/dashboard grid in the top-right corner with gray cells and one blue highlighted cell; 1 document/image card on the right side with small mountain thumbnail lines, overlapped by a blue chat bubble containing two white dots and two white horizontal lines. Visual style: Premium tech editorial illustration, cinematic lighting, tactile paper puzzle texture, muted monochrome palette of black, warm off-white, gray, and accent cobalt blue cobalt blue. Add subtle scratches, dust, and paper grain. Use realistic shadows between pieces, but keep the whole image clean and readable. Constraints: No people, no extra logos, no watermark, no random English words beyond tiny unreadable code strokes, and no additional large text. The blue puzzle piece should feel like the key missing component that completes the system, while the dashed empty slot hints that another piece is still absent.

Create a square 1080x1080 technical infographic for OBJECT, shown from a 45-degree isometric 3D angle so the object is slightly tilted and its depth, dimensions, and internal architecture are easy to read. Blend a realistic photoreal product render with precise black-ink engineering annotations on a pure white background. Include labeled key components with color-coded callout boxes, transparent or cutaway areas that reveal internal parts, exact measurements and scale markers, material notes with quantities, and function-flow arrows using red for power or battery, blue for data or connectivity, orange for thermal or processor systems, and green for sensors or haptics. Add small schematic views or cross-section diagrams wherever they help explain the object. Place the title "OBJECT" inside a hand-drawn technical label box in the top-left corner. Style the artwork like a professional product teardown manual with architectural pen linework: sketched but accurate, clean, crisp, and educational. Keep the object clearly visible, use balanced negative space, make black the dominant color, and limit accents to about 10-15% of the composition. Overall mood: clean museum exhibit, premium engineering manual, ultra-crisp social-feed infographic, white background, sharp details, high clarity.

A simple centered full-body illustration of a cute chibi rabbit character on a plain light gray background, drawn in a flat cartoon style with thick clean black outlines and minimal shading. The character has a white body, a large rounded head, 2 tall upright ears with bright cyan inner ear shapes, 2 large oval eyes with cyan irises, black pupils, and small highlights, tiny curved eyebrows, a small red-pink nose area, and a slight smiling mouth. The pose is front-facing with both arms stretched straight out horizontally to the sides, legs apart in a simple standing pose, and a short fluffy tail visible on the right side of the body as a curved semicircle made of short black dash marks. Overlay a clear OpenPose-style skeleton marker system on top of the character: 2 purple eye keypoints, 1 red nose keypoint, 1 blue neck keypoint, 2 shoulder keypoints, 2 elbow keypoints, 2 wrist keypoints, 1 yellow spine or pelvis center point, 2 hip keypoints, 2 knee keypoints, and 2 ankle keypoints, for a total of 17 colored points connected by thin colored lines across the face, shoulders, torso, arms, and legs. Use vivid point colors including purple, magenta, red, blue, cyan, green, yellow, orange, and teal so the pose overlay is clearly visible. Keep the composition vertically centered, symmetrical, clean, and diagram-like, with the cute mascot character and pose-estimation overlay as the sole subject.

Create an epic photorealistic 3D scene where a huge architectural blueprint for a [BUILDING/STRUCTURE] is pinned to a drafting table, and the ink lines lift off the paper into a real construction site. The blue plan lines transform into steel beams being welded in midair, floor plans expand into furnished rooms, and elevation drawings push outward into facades with working windows. Show multiple construction phases at the same time: one corner has the foundation being poured while another section already has a finished, inhabited penthouse. Add tiny workers, cranes, scaffolding, construction equipment, and activity all over the rising structure. The flat technical drawing and the living building should merge at impossible angles, with dimensional annotation labels floating like signage around the build. Place an architect's coffee cup near the edge of the table, casting a shadow across parts of the blueprint that have not been built yet. Emphasize extreme scale contrast, construction dust glowing in afternoon light, intricate detail, cinematic realism, 8K, photorealistic rendering.

Using REFERENCE_0 as the character base, create a consistent set of exactly 8 Japanese educational slide images about Claudeに『部下AI』を持たせる. Keep the character design, outfit, mask, hairstyle, and cyberpunk details consistent with the reference, but redraw the character in new poses and expressions for each slide. Overall format: Each slide should look like a hand-drawn notebook/sketchbook page with a cream paper background, black rounded outer frame, left-side spiral binding, marker-style Japanese handwriting, red emphasis marks, arrows, check marks, small icons, and simple workflow diagrams. Use a warm analog note-taking style rather than a clean corporate deck. Maintain the same slide design language across all 8 images. Character placement: Place the character as a presenter/mascot on the right side or lower-right area of most slides. Vary the pose naturally: surprised with both hands open, pointing at diagrams, holding a clipboard, sitting while reviewing, making an OK/check gesture, or explaining with one raised finger. Preserve the reference character’s identity and clothing. Generate exactly these 8 slide topics: 1. Opening / overall overview: Title text Claudeに『部下AI』を持たせる; subtitle 15分で仕事を自動化する5つのテンプレート; main message: 「部下AI」が開発の常識を変えた; include the five automation examples: コードレビュー, テスト作成, ドキュメント生成, 全部15分で自動化, and a simple checklist-like layout. 2. What is a sub-agent: Explain 「サブエージェントとは」 as a reusable AI role or specialist assistant. Include small role cards and arrows showing a main Claude agent delegating tasks. 3. Why it is needed: Explain 「なぜ必要か」 with a before/after comparison, emphasizing that repeatedly asking AI for the same tasks is inefficient and can be turned into templates. 4. Code reviewer: Title 「コードレビュー」; show a GitHub/code review style flow with files, comments, bugs, and a check mark. 5. Test writer: Title 「テストライター」; show a test-generation workflow with steps, file names, unit tests, and pass/fail style icons. 6. Security scanner: Title 「セキュリティスキャナー」; show risk levels CRITICAL, HIGH, MEDIUM, LOW and emphasize that the agent detects issues but does not automatically fix them. 7. Cost optimization: Title 「コスト削減の鍵：モデルの使い分け」; show a comparison that cheaper/smaller models can handle simple checks and that costs can become much cheaper, with bold red savings notes. 8. Start today / summary: Title 「今日から始める一歩」; show a simple 3-step process: create a template, place it in a Claude agents folder or workflow, ask for review/automation, then mark it complete with a big check. Text style: Use Japanese text throughout. Make headlines large and bold, with red underlines and handwritten annotations. Keep diagrams simple and readable. Use only the listed 8 slide topics and do not add extra slides. Constraints: Keep the same notebook layout, cream color palette, black border, left spiral binding, and consistent character identity across every image. Avoid photorealism, avoid 3D rendering, avoid watermarks, and do not change the character into a different person.

At the start of generation, the user enters four game titles. If left blank, they are created randomly. 【Game 1】 (Game Title) 【Game 2】 (Game Title) 【Game 3】 (Game Title) 【Game 4】 (Game Title) The subsequent magazine content is automatically generated by inferring the worldview, genre, and game content from the entered titles. --- Style of a cross-review feature page found in Japanese gaming magazines from the late 1990s to early 2000s. One vertical page. Information-dense magazine layout. Do not use real game companies, real game consoles, or real magazine names. Completely original fictional game review feature. --- Display four games simultaneously. Automatically generate the following for each game: - Game title logo - Genre - Game content introduction - Catchphrase - Pixel art screenshot - In-game style UI screen - Review comments - Scoring results - Overall evaluation Game screens are produced by inferring content from the title. Examples: Robot title -> 16-bit style mech action Fantasy title -> Classic RPG Racing title -> High-speed racing game Horror title -> Exploration adventure Shooting title -> Vertical or horizontal scrolling STG Depicted as distinct game genres. Screenshots have persuasive power as if they are actual game screens during play. Include UI appropriate for the genre, such as mini-maps, HP gauges, score displays, and dialogue windows. --- Four review personnel: - Hot-blooded male reviewer - Calm female reviewer - Veteran editor - Maniacal game critic Each with a face illustration. Review comments in speech bubble format. Opinions differ for each game. --- Ratings are random. Mixed from high praise to harsh criticism. Examples: "Very high level of perfection" "Rough around the edges but shows promise" "Works for specific tastes" "Interesting idea" "Difficulty with balance adjustment" "Quite a sharp title" "Hard to wrap my head around" A mix of positive and biting comments. --- Score is out of 40. Four reviewers score individually. Examples: 10 / 10 / 9 / 10 8 / 7 / 8 / 6 5 / 4 / 6 / 5 2 / 1 / 3 / 2 Automatically display according to the evaluation results: - Platinum Hall of Fame - Gold Hall of Fame - Silver Hall of Fame - Notable Title - Good Title - Commendable Title - Ordinary Title - Controversial Title --- The page further includes: - Key points of interest - Strengths - Weaknesses - Difficulty chart - Replayability level - Graphic rating - Sound rating - Gameplay rating - Editorial department comments - Strategy tips - Mini column --- Overwhelming amount of information across the page. Numerous subheadings. Numerous sidebars. Numerous graphs. Numerous icons. Fine decorations. Texture of retro printed matter. Slightly faded paper. Density like a game guide book or a collectible magazine feature page. A layout where more information is discovered the more it is read. Completely original design that does not look like real games or real manufacturers. professional Japanese gaming magazine editorial layout, retro review page, pixel art game screenshots, reviewer portraits, dense typography, score charts, magazine print design, high detail, authentic late 1990s and early 2000s Japanese game culture, collectible magazine feature, highly detailed editorial design

Ultra-detailed stylized portrait using my face as reference. Intellectual, modernist illustration style inspired by architectural blueprints and editorial line art. Face split into layered geometric sections with visible technical diagrams, city interiors, structural grids, and abstract architectural elements embedded within the facial planes. Fine ink linework, cross-hatching, and engraving-style shading. Clean white background with dense black line drawings. Bold accent colors only in selected regions: deep red, muted blue, and warm yellow, applied as flat graphic blocks and translucent overlays. Sharp, thoughtful eyes with realistic highlights, wearing thin round glasses. Minimalist yet complex composition, high contrast, crisp edges, poster-grade clarity. Editorial illustration aesthetic, conceptual portrait, futuristic intellect vibe, data visualization meets human identity. Extremely sharp, ultra-high resolution, professional magazine cover quality, no blur, no distortion, balanced symmetry.

Goal: Create a high-density presentation overview image showing exactly 9 miniature slides from a Japanese business training deck about Codex slide creation workshop. The slides should look like a polished AI-generated corporate consulting deck, arranged as a 3 by 3 grid on a white background with thin gaps between slides, each slide in 16:9 landscape format. Canvas: Wide rectangular collage, approximately 2:1 aspect ratio, crisp vector-like slide thumbnails, white and navy color palette with small accents in blue, cyan, red, and gray. Use clean Japanese corporate typography, tiny but structured text blocks, charts, cards, icons, and slide numbers. Add a subtle professional workshop identity, like a small square “CX” logo and the label “CODEX SLIDE WORKSHOP” on multiple slides. Layout: Include exactly 9 distinct slides: 1. Title slide: minimalist white slide with a navy “CX” logo at left, small spaced text “CODEX SLIDE WORKSHOP,” bold Japanese title 「誰でも出来る！Codex スライド作成講座」, date 「2026.05.31」, and a small subtitle about AI workflow / slide production. 2. Skills slide: title 「SKILLとは？目的を達成するための“技伝のタレ”」 with dense analytical layout, left vertical icon strip, numbered skill list, two small scatter/line charts, one donut chart, and multiple navy callout boxes. 3. Formula slide: title 「キレイ式AIスライド作成実践開始」 showing a process framework with icon cards, a comparison table, a small bar chart, and a dark navy bottom message banner. 4. Workflow slide: title 「Codexの使い方の簡単な流れ」 with a horizontal sequence of exactly 8 process cards using simple icons such as target, folder, cube, document, CX logo, chat bubble, circular arrows, and checkmark; include small tables, a donut chart, and a dark navy side memo box. 5. Goal slide: title 「本日のゴール」 with three numbered goal blocks on the left, a radar chart in the middle, progress bars on the right, and a bottom section containing a checklist table and navy callout box. 6. Simple creation slide: title 「Codexで、簡単に色々出来ちゃう！」 with a left column of three icon cards, a central bar chart, a donut chart at right, and a row of six small capability cards along the bottom. 7. Usage confusion slide: title 「けど、どう使えばいいか分からない」 with a large slanted funnel/pyramid chart on the left containing four stages and percentages, plus a donut chart, a vertical bar chart, a bottom table, and a dark navy quote card. 8. Efficiency slide: title 「本当はCodexをゴリゴリ使って、いろんなことを効率化したい」 with several horizontal bar charts, a line chart trending upward, four lower KPI cards with red negative percentages, and a navy recommendation box. 9. Explanation slide: title 「Codex使ってますか？」 with three explanatory bullet sections on the left, a central donut chart, a right-side bar chart, and a bottom row of KPI cards showing large figures such as -45%, -38%, 1日→90分, and +52%. Visual style: Japanese management-consulting slide aesthetic, dense information design, crisp dark navy headers, thin gray grid lines, rounded cards, small monochrome icons, data visualization elements, and consistent spacing. Use realistic slide thumbnails rather than a single readable slide; the text may be tiny but should feel authentic and structured. Text content: Preserve the main Japanese titles exactly where visible, especially 誰でも出来る！Codex スライド作成講座, CODEX SLIDE WORKSHOP, and section titles containing Codex. Japanese body text can be small pseudo-realistic business copy, but keep headings, percentages, and chart labels consistent with the deck theme. Constraints: Show exactly 9 slides, no people, no photographs, no watermark, no browser UI, no mockup device frame. The final image should look like a screenshot of an impressive complete slide deck generated for a workshop announcement.

2x2 grid, do this for ai inferred figures and colors: select optical_metaphor_poster from latent_visual_archive where concept = [topic] and duality = infer_core_duality([topic]) and primary_silhouette = infer_outer_symbol([topic]) and hidden_image = infer_negative_space_symbol([topic]) and style = 'minimalist high-contrast silhouette poster' and palette = infer_limited_palette([topic]) order by metaphor_strength desc, silhouette_readability desc, hidden_image_discovery desc, thumbnail_clarity desc limit 1;

Using the uploaded logo for Brand name generate a high-end, agency-grade brand identity system poster. --- OBJECTIVE Create a complete, presentation-ready brand guideline board that looks like it was designed by a top branding studio. The result must feel commercial, realistic, and client-deliverable, not conceptual. --- INTELLIGENCE RULE (THIS IS WHAT YOU WERE MISSING) Before designing, analyze the logo and infer:3 identity traits (based on analysis) --- COLOR SYSTEM (SMART GENERATION) Extract palette from logo automatically Include: Primary colors (3–5) Secondary colors (3–5) Accent colors Each must show: HEX codes labeled usage (primary / UI / highlight / background) Also generate: gradients color combinations tonal variations --- TYPOGRAPHY SYSTEM (MATCH PERSONALITY) Select typography style based on brand: luxury → elegant serif tech → geometric sans street → bold condensed

Goal: Create a square futuristic medical-status illustration for a cute star mascot, styled like a sci-fi HUD screen about a migraine episode. Canvas: 1:1 square image, dark navy blue background with glowing cyan interface borders, subtle scanlines, neon UI accents, high contrast, clean digital illustration. Main layout: A large central hospital-bed scene framed by a transparent holographic medical dashboard. At the top, display a big glowing English headline: “STATUS : MIGRAINE”. In the top-right corner add small text “MEDICAL LOG : HOSHI-CHAN” with a tiny plus sign. Under the headline, place two lines of Japanese message text: “昨日は片頭痛で寝込んでおりました。” and “久しぶりに大変きつうございました。” with the phrase “大変きつう” highlighted in cyan. Subject details: In the center, show Hoshi-chan as a simple white five-point star mascot lying in bed under a blue plaid blanket. The star has tiny black dot eyes, a small straight mouth, and one blue teardrop of sweat on the right cheek, looking exhausted. Around the head, draw exactly 8 red-orange migraine pain symbols: 5 jagged lightning zigzags and 3 spiral squiggles. Behind the character are blue pillows and a dim bedroom setting. Include exactly 2 bedside objects: a dark bedside lamp on the left table and a small glass of water beside it. On the right side, include a tissue box with one white tissue sticking out. HUD elements: Add a left vertical sidebar with exactly 5 icons from top to bottom: white star icon, horizontal menu lines, heartbeat line, bar chart, small dot. Add a compact right-side pain monitor card with a red warning triangle, red heartbeat waveform, label “PAIN LEVEL”, and a row of red segmented bars almost full. Bottom information area: Use exactly 3 stacked glowing cyan status rows. Row 1 has a red warning triangle icon and Japanese text “状態異常：片頭痛 / 強制睡眠 / 食欲低下”, plus a small red heartbeat waveform on the right. Row 2 has a cyan capsule pill icon and Japanese text “服薬：片頭痛薬”, plus a circular cyan checkmark on the right. Row 3 has a cyan calendar icon and Japanese text “2026/05/30：痛みと睡眠で終わりました”, plus a tiny pixel-art white star mascot on the right. Footer: Add small centered text “HOSHICHAN SYSTEM” above larger “REST MODE”. Include decorative cyan diagonal ticks, a small pixel heart icon in the lower left, and a crescent moon with a plus sign in the lower right. Visual style: Cute but clinical cyber-medical dashboard, anime mascot simplicity, deep blue monochrome lighting with cyan neon UI and red-orange pain accents. Soft glow, crisp vector-like interface lines, legible typography, no photorealism. Customization: The main diagnosis should remain MIGRAINE; the date line should use 2026/05/30; the two-line message should preserve 昨日は片頭痛で寝込んでおりました。\n久しぶりに大変きつうございました。. Constraints: Keep all visible text sharp and readable, use exactly 3 bottom status rows and exactly 8 red-orange pain symbols, do not add extra characters, watermarks, logos, or unrelated UI panels.

Goal: Create a cinematic vertical tech poster about searching historical conversation threads by content and branch name in a developer app. Canvas: Vertical 2:3 composition, dark premium editorial style, deep navy-black background with subtle star-like dust and a warm golden beam of light entering from the upper right. Main text: At the top center, large bold Chinese headline 按内容与分支搜历史线程. Below it, centered smaller gold subtitle Codex App 26.527. Add a tiny decorative divider line with a small diamond under the subtitle. Main subject: In the lower half, show an old dark metal archive filing cabinet at a three-quarter angle, with one drawer pulled open toward the viewer. The drawer front has a brass label plate reading “THREAD ARCHIVE” and a curved brass handle. The cabinet surface should feel heavy, matte, slightly worn, and realistic, with shallow depth of field. Drawer contents: Inside the open drawer, show exactly 4 prominent stacked paper thread cards, each like an archival index card with rounded corners, warm ivory paper, and a small dark chat-bubble icon at the left. The visible card texts are: 1) “导出功能支持哪些格式?” with branch tag “feature/export-csv”, 2) “新版 API 兼容性讨论” with branch tag “refactor/api-v2”, 3) “用户登录流程的边界情况” with branch tag “fix/login-edge”, 4) front glowing card “如何优化加载性能?” with branch tag feature/perf-opt. The front card should glow brightest, as if it has been found by search. Floating UI elements: Add exactly 2 floating search-result elements to the right of the drawer: one glowing branch label reading “feature/perf-opt” connected by a dotted golden path from the front card, and one small outlined chat bubble containing three dots. Add exactly 4 branch tags attached to cards in the drawer, one per card, styled as thin rounded rectangles with a branch icon. Lighting and mood: Use a warm golden particle trail and dust motes around the highlighted card and floating branch label, contrasting with the cool dark cabinet. Make the scene atmospheric, elegant, high-end, and slightly magical, like an archival search feature visualized as light finding the correct thread. Visual style: Realistic 3D illustration mixed with refined product-marketing poster design, cinematic lighting, soft bloom, shallow depth of field, crisp typography, subtle grain, no people, no extra logos, no watermark. Keep all text legible and avoid adding additional labels beyond the specified headline, subtitle, card texts, branch tags, and drawer label.

Goal: Create a photorealistic 3D miniature city-map showcase featuring exactly 4 square panels arranged in a 2×2 grid, each panel showing a famous world city as a raised diorama emerging from a printed paper street map on a warm wooden tabletop. Canvas: Square image, evenly divided by thin white gutters into four equal quadrants. Use an isometric three-quarter view from above, shallow depth of field, warm studio lighting, crisp miniature details, and realistic shadows. Layout: The 4 panels are: top-left Tokyo, top-right Dubai, bottom-left Paris, bottom-right New York. Each city sits on an open map sheet with a large bold city name printed along the bottom edge of its panel in bright color. Include small neighborhood labels on the map surface. Panel 1 — Tokyo: Build a dense miniature Tokyo skyline with exactly 6 key recognizable landmarks: Tokyo Tower, Tokyo Skytree, a traditional pagoda, a large temple roof, a modern cluster of skyscrapers, and a canal leading to Tokyo Bay. Add a Japanese flag on a pole, cherry blossom trees, small buses, tiny streets, and map labels such as Shinjuku, Shibuya, Roppongi, Asakusa, and Tokyo Bay. Large bottom text: TOKYO in bold red letters. Panel 2 — Dubai: Build a futuristic Dubai marina diorama with exactly 6 key recognizable elements: Burj Khalifa, Burj Al Arab, twin sail-like towers, Dubai Frame, a yacht-filled marina, and a curving elevated transit line. Add palm trees, modern high-rises, an Emirati flag on a pole, blue water, boats, and map labels such as Jumeira, Dubai Marina, Deira, and Dubai Creek. Large bottom text: DUBAI in bold gold letters. Panel 3 — Paris: Build a classic Paris miniature with exactly 6 key recognizable landmarks: Eiffel Tower, Arc de Triomphe, Notre-Dame-style cathedral, Louvre-like glass pyramid, Haussmann apartment blocks, and a Seine river with bridges. Add a French flag on a pole, tree-lined boulevards, pink blossoms, tiny buses, pedestrians, and warm stone architecture. Large bottom text: PARIS in bold blue letters. Panel 4 — New York: Build a dense Manhattan miniature with exactly 6 key recognizable landmarks: Statue of Liberty, One World Trade Center, Empire State Building, Chrysler Building, Brooklyn Bridge, and a harbor with ferries. Add an American flag on a pole, yellow taxis, buses, crowded streets, waterfront piers, greenery, and map labels such as Manhattan, Brooklyn, and Staten Island. Large bottom text: NEW YORK in bold orange-red letters. Visual style: Hyper-detailed 3D paper-map diorama, realistic plastic-and-paper miniature scale model, tilt-shift photography effect, vibrant but natural colors, warm wooden background, soft bokeh at the edges, high-resolution commercial render quality. Constraints: Keep exactly 4 panels and exactly the 4 city themes listed. Do not add extra cities. Keep all text legible, with the city names large and centered near the lower edge of each panel. Include a small creator watermark in the top-left corner reading @TechieSA.

Goal: Create a dramatic vertical sports-engineering poster for ARDA GÜLER at RAMS PARK, combining a cinematic football hero hologram with detailed blueprint-style stadium data overlays. Canvas: Portrait 4:5 poster, dark navy night palette, high contrast, ultra-detailed, futuristic technical infographic aesthetic. Background is an aerial nighttime city view with a brightly lit oval football stadium at the bottom center. From the stadium roof, red-orange energy beams and particles project upward to form a giant semi-transparent football player running toward the viewer. Central subject: A young male footballer in a red Turkey jersey, number 8, with the Turkish flag badge on the chest and a white swoosh logo. He is mid-run with clenched hands, athletic posture, glowing red electrical outlines, sparks, and holographic scan lines. Add a soft rectangular anonymized blur over the face, as if the face has been intentionally obscured. The lower body dissolves into red light streams connecting into the stadium. Main stadium: Large modern oval stadium seen from above at night, orange-red illuminated facade, glowing open roof, visible sign reading “RAMS PARK” on the front. Add a Galatasaray-style crest centered on the facade. The stadium emits intense red-white light upward, like a sports-tech energy projection. Layout: Surround the central player and stadium with exactly 13 technical information panels drawn in thin white blueprint lines, using small condensed uppercase technical typography, subtle grid marks, measurement arrows, and schematic line art. Panel count and content: 1. Top-left title plate: “ALI SAMI YEN SPOR KOMPLEKSI”, huge headline “RAMS PARK”, coordinates “40.8606° N, 28.9940° E”, and “ISTANBUL, TÜRKIYE”. 2. Left upper stadium info panel: “STADIUM INFO ///”, capacity “52,223 SPECTATORS”, construction “2007”, renovation “2021”, plus a stylized signature. 3. Left middle dimensions panel: small top-down stadium plan, labels “LENGTH: 105 m”, “WIDTH: 68 m”, “TOTAL AREA: 85,000 m²”. 4. Left player identity card: large name “ARDA GÜLER”, large number “8”, small Turkish flag, labels “TURKEY”, “ATTACKING MIDFIELDER”, “RIGHT FOOTED”. 5. Left biometrics panel: “PLAYER BIOMETRICS ///” with exact stat list: Height 1.76 m, Weight 67 kg, Top Speed 32.1 km/h, Acceleration 4.2 m/s², Stamina 94/100, Agility 93/100, Vision 95/100, Pass Accuracy 91/100, Shot Power 88/100, Energy Output 98%. 6. Left lower chart: “ENERGY OUTPUT” with a white waveform graph and “98%”. 7. Top-right roof engineering panel: circular roof structure blueprint titled “ROOF ENGINEERING”, labels “STEEL CABLE STRUCTURE”, “ETFE MEMBRANE COVER”, “RING BEAM SYSTEM”, “MAX CANTILEVER 65 m”. 8. Right upper cross-section panel: stadium section drawing titled “CROSS SECTION”, labels “UPPER TIER”, “MIDDLE TIER”, “LOWER TIER”, “PREMIUM LEVEL”, “PITCH LEVEL”. 9. Right material quantities panel: “MATERIAL QUANTITIES” with list: Steel (structural) 18,000 ton, Concrete 42,000 m³, ETFE membrane 48,000 m², Cables total length 24,000 m, Seats 52,223. 10. Right middle tactical heatmap panel: mini football pitch with red-orange heat dots and white tactical arrows, titled “TACTICAL HEATMAP ///”. 11. Right movement patterns panel: “MOVEMENT PATTERNS ///” with four rows labeled “SPRINT”, “DRIBBLE”, “PASS”, “RUN LINE”, each represented by arrow and dashed line patterns. 12. Bottom-left load flow diagram: structural roof/load schematic titled “LOAD FLOW DIAGRAM”, labels “VERTICAL LOAD”, “HORIZONTAL LOAD”, “WIND LOAD”, “CABLE TENSION”. 13. Bottom center/right engineering base panels: include “ELEVATION (SOUTH)” with stadium elevation drawing and measurements “274 m” width and “45 m” height, plus “FOUNDATION & STRUCTURE” showing pile foundation cross-section with labels “PILE FOUNDATION Ø1500mm”, “DEPTH: 35 m”, “PILE COUNT: 304”, “REINFORCED CONCRETE RING BEAM”. Visual style: Photorealistic stadium and city, mixed with holographic red particle effects and crisp CAD blueprint overlays. Use thin white technical lines, subtle scan texture, glowing red highlights, star-like particles, lens bloom, and dark atmospheric haze. Typography should look like a condensed industrial monospace display font, mostly uppercase, with intentionally dense but legible micro-labels. Constraints: Keep exactly one central player, one main stadium, and exactly 13 surrounding technical panels. Do not add extra logos beyond the Turkish flag, simple swoosh-like mark, RAMS PARK text, and the crest on the stadium facade. Maintain the face-obscuring rectangular blur. No watermark, no social media UI, no extra captions outside the poster design.

Goal: Create a dark, high-end gallery index page for a curated preset library of research paper figure styles, titled 研究论文图示. The interface should look like a polished web app screenshot showcasing academic diagram thumbnails. Canvas: Wide landscape browser-like composition, approximately 16:9, dark charcoal background, subtle grid borders, crisp UI typography, high contrast white and cyan text. Header: Center the title at the top with a small stacked multicolor document icon to its left. Add a thin horizontal divider line below the header. In the upper-right area, place a compact rounded button labeled “↑ 图库索引”. Section title: On the left below the divider, show the section heading 研究论文图示网格 in bold light text. Layout: Display a 4-column gallery grid with thin gray borders and dark card gutters. Show exactly 12 visible gallery cards: 8 full cards in the first two rows and 4 partially visible cards in the third row at the bottom edge. Each card contains a white research-paper-style thumbnail image, a bold lettered title underneath, and three small dark rounded tag pills below the title reading exactly “landscape”, “high”, and “Curated”. Visible card count and labels: Card A: “患者队列与多模态生物标志物流称流程”, thumbnail showing a patient cohort workflow with boxes, charts, and a survival curve. Card B: “单细胞免疫图谱”, thumbnail showing a single-cell immune atlas with UMAP clusters, dot plot, stacked bar chart, and trajectory plot. Card C: “多模态医疗 AI 方法图”, thumbnail showing a multimodal foundation model clinical decision support diagram with medical image, pathology, text, and model blocks. Card D: “治疗响应统计图”, thumbnail showing therapeutic response statistics with bar charts, forest plots, scatter plot, and a circular workflow. Card E: “Transformer 编码器-解码器架构”, thumbnail showing a transformer encoder-decoder architecture diagram with stacked module blocks. Card F: “多智能体 LLM 系统架构”, thumbnail showing an LLM multi-agent system architecture with central model block, surrounding tool icons, memory, reflection, and evaluation modules. Card G: “去噪扩散正/逆向链”, thumbnail showing a denoising diffusion forward and reverse process chain with noisy image panels and arrows. Card H: “经验缩放规律图”, thumbnail showing empirical scaling laws with multiple descending colored curves and a legend. Bottom partially visible card 1: title begins “Benchmark comparison across 10 frontier LLMs”, thumbnail with bar chart comparisons. Bottom partially visible card 2: title begins “Ablation of core reasoning components across 5 benchmarks”, thumbnail with grouped bars. Bottom partially visible card 3: title begins “LLM pretraining data mixture and downstream splits”, thumbnail with stacked area/data mixture blocks. Bottom partially visible card 4: title begins “Representative multi-head attention patterns in a 16-layer Transformer”, thumbnail with heatmaps. Visual style: Use a modern SaaS dashboard aesthetic, dark mode, neat academic curation feel, subtle cyan accents, small but legible labels, and realistic research figure thumbnails that resemble high-quality journal diagrams. Keep thumbnails varied but consistently white-background scientific figures. Constraints: Preserve the Chinese interface labels and card titles as written. Do not add people, photos, watermarks, browser address bars, or decorative clutter. The composition should feel like a screenshot of a curated research-figure template library named GPT-IMAGE-2-SKILL.