原创栏目企划 / 尚未实测与拍摄 / 已完成策划及视觉预演

内置AI生图：最终提示词与资产记录

生成方式：内置image_gen工具，2026-10-21。没有调用CLI/API后备、没有程序绘图、没有新SVG。全部为概念预演，非真实人物、实测现场、软件录屏或工具输出。图中人物是虚构概念角色，不能作为真实采访对象。

最终文件：assets/hero-moodboard.png、assets/storyboard-8frames.png、assets/cover-a-v2.png、assets/cover-b.png；cover-a.png为保留的首版概念。生图后原样复制进项目，图板在网页按格裁切，未用程序改图。原始生成文件同时保留在Codex generated_images目录。

01 / hero-moodboard.png

Use case: ads-marketing. Asset type: landscape editorial hero and moodboard for an ORIGINAL PLANNED Chinese short-video show about testing AI on a real job. This is concept previsualization, no real experiment footage. Create a sophisticated cinematic still life in a modest Shenzhen video editor's studio: an orange task folder, large dark monitor showing a deliberately abstract video editing timeline (no readable software brand), a simple mechanical countdown clock with its face mostly unmarked, a paper assessment form with unfilled checkboxes, hand about to press a keyboard. Empty chair as the AI 'job applicant'; no robot, no neon, no hologram, no money or scores. Composition: wide landscape 16:9, documentary photographic collage look, left 55% desk and monitor, right 45% arranged paper details and small material/palette chips, excellent negative space top left for HTML headline overlay. Restrained palette: paper #F2F0E9, charcoal #202522, burnt orange #EE7544 and acid yellow #D9EB84. Natural warm window light, tactile paper, subtle film texture, design editorial quality, young internet task show, quietly humorous job audition. No Chinese or English text, no logos, no watermarks, no data that implies a test result, no SVG look. This image must clearly feel like a crafted visual concept rather than claimed real documentary footage. Generate a polished bitmap landscape image.

02 / storyboard-8frames.png

Use case: illustration-story. Asset type: EIGHT FRAME STORYBOARD CONTACT SHEET for original planned Chinese vertical short video 'AI JOB AUDITION: VIDEO EDITOR'. Create ONE landscape sheet with EXACTLY FOUR columns by TWO rows. Every one of the eight panels must be a precise 9:16 upright composition; the sheet overall ratio therefore 9:8 with slim even charcoal gutters. Numbering outside imagery is not required; absolutely no captions or readable UI text. Sophisticated cinematic documentary illustration style with photographic textures, clearly conceptual previsualization, consistent restrained burnt orange #EE7544, charcoal #202522, paper #F2F0E9 and accent acid yellow #D9EB84. Same ordinary video editing desktop and same hands throughout, no robot, no neon, no logo. Panel 1: close-up orange task folder being slid onto dark desk, monitor in background. Panel 2: medium shot of young adult Chinese video editor, three-quarter back, seated at ordinary desk, a blank orange task brief clipped to monitor. Panel 3: overhead clean layout of eight small generic video thumbnails on paper strips and blank rights checklist, no ticks. Panel 4: over-shoulder monitor with abstract text input and upload tiles, hand selecting source footage. Panel 5: macro mechanical timer beside keyboard, display unmarked, hand hovering to start. Panel 6: vertical monitor showing generic abstract 30-second edit timeline, orange clip blocks with an indistinct preview image, no software branding, no claimed result. Panel 7: hands comparing two frame screenshots with an unfilled checklist, neutral assessment, no red/green score or verdict. Panel 8: full desk shot with orange envelope and unfilled final assessment card, human interviewer hand writing question mark only. Keep vital action inside central 70 percent of each panel. Each frame has room in upper 15 percent and lower 20 percent for HTML titles, no gibberish text. This is a premium documentary task-program storyboard, not real test evidence. No scores, no invented timings, no pass/fail stamps. Sharp, readable actions, editorial design, tactile natural light.

03 / cover-a.png

Use case: ads-marketing. Asset type: 9:16 vertical video COVER concept for original Chinese AI task challenge show. Main exact Chinese headline in huge confident black condensed sans typography at top half: “面试AI剪辑师”. Small series wordmark at very top: “AI上岗考核”. Bottom modest line: “别急着吹，先把活干了。” Visual idea: editorial still life, orange job audition task folder slapped on real dark editing desk; empty office chair behind a monitor with generic blurred video timeline. An off-screen human hand places a plain paper task brief. Photo collage with natural warm light and tactile texture, premium task-program graphics, not cartoon. Bold orange #EE7544 upper half background, charcoal #202522 and paper #F2F0E9, one acid yellow #D9EB84 rectangular blank tag. Bottom micro label exact text: “概念预演 · 尚未实测”. No pass/fail stamp, no result, no score, no fake brand interface, no robot, no neon, no sci-fi grid, no additional words. Composition optimized for phone and vertical 9:16 crop, big readable typography, strong air, visual mass centered, keep rightmost 12 percent and lowest 13 percent low detail for platform UI. Premium restrained internet tech program aesthetic.

04 / cover-b.png

Use case: ads-marketing. Asset type: SECOND DIFFERENT 9:16 vertical video cover concept for original planned Chinese AI task challenge show. Exact series wordmark small at top “AI上岗考核”. Exact huge stacked Chinese title in top half “这份粗剪” / “能交差吗？” in paper-white sans serif; the question mark is orange. Exact small eyebrow “面试AI剪辑师”. Cover composition: mostly charcoal #202522 background with tactile off-white horizontal editing timeline strips passing diagonally across middle; large hands hold a plain paper assessment sheet with unfilled square checkboxes over a real editing desk. Timeline generic and unbranded, few orange #EE7544 clip blocks; acid yellow #D9EB84 small plain tag with exact text “待验收”. No completed score, no percentage, no pass/fail, no smiley, no robot, no neon, no fake test result. Photo collage editorial style, warm natural desk light, premium restrained work-task challenge identity, quietly humorous. Clearly a crafted concept visual, not claimed real footage. Bottom micro label “概念预演 · 尚未实测”. Make layout highly readable on phone with strong whitespace, visual action in central safe region and lowest 13 percent clear. Keep all spelling exact, no other text. Vertical aspect 9:16.

视觉QA与使用限制

已检查：4项资产均已生成并进入项目；两款封面标题可读；任务图板各格一致采用竖屏；图中无伪造分数和通过结论。界面只是抽象概念，不对应指定工具；图板中的素材缩略图也不是已经获权的正式测试输入。正式开机须重新拍摄、取得授权并替换。

封面A首版的酸黄留白标签在最终v2中通过内置生图编辑补为“任务待实测”，首版保留以便追溯。网页把任务条件排成真实可读文字，而不依赖生成图中小字。生成图不作为事实证据。

05 / cover-a-v2.png · 最终修订

内置image_gen编辑，输入为本项目已查看的cover-a.png。仅修订留白标签，保留首版文件。

Use case: precise-object-edit. Edit target: the provided 9:16 original concept video cover. Change ONLY the currently empty acid-yellow rectangular paper tag in the upper-middle of the cover. Add exact black Chinese text centered on that tag: “任务待实测”. Use the same angle and perspective as the yellow rectangle, confident plain sans Chinese, sharply readable. Keep EVERYTHING ELSE unchanged: exact headline “面试AI剪辑师”, series label “AI上岗考核”, slogan, bottom concept-state label, 9:16 composition, all orange paper, desk, empty chair, monitor, lighting, color, hands, typography. Do not add a result or numerical score. This remains concept previsualization and has no test data. No new objects, no crop.