Toggle Sidebar

技能广场我的技能

上传生成

全部(39)图片(26)视频(4)音频(2)3D(2)工具(5)

仅未添加

查看详情添加技能

y2k-camcorder-look

Apply a fixed "Y2K camcorder" photo filter: an uploaded photo (or a text subject brief) is rendered as a cheap 2003 consumer DV camcorder snapshot — timestamp overlay, blown flash, low-res grain, faint VHS tracking, off-center framing. One image, no choices, no questions. Use when the user uploads a photo and asks for retro / vintage / VHS / camcorder / old-camera / 2000s styling, or sends an "old camcorder photo" subject brief. Opposite register to pastel-y2k-fashion-portrait (candy-pop hyperreal) and Polaroid skills (instant print) — do not bend toward either. Routes through image-executor via image_create (T2I / I2I).

查看详情添加技能

vintage-patent

Render anything as an authentic late-1800s USPTO patent document — aged ivory paper, fountain-pen marginalia, embossed seal, multi-figure technical drawings. Turns any invention, joke, or uploaded reference image into a "recovered archival filing." Use when the user asks to patent an invention, make a vintage patent-style image or print, convert a photo/drawing into a patent figure, or mock-patent a product/idea as an 1880s filing.

查看详情添加技能

video-subtitle-extractor

视频字幕提取器，上传本地视频，自动提取字幕并生成转录稿。

查看详情添加技能

video-generation-skill

为营销视频设计脚本、分镜和剪辑方案，支持短视频和长视频多平台内容规划

查看详情添加技能

tvc-storyboard

创建 TVC 广告分镜。用于当用户想要商业影片概念、TVC 脚本、分镜镜头表、3x3 分镜图、活动视觉方向，或基于产品简报、品牌简报、脚本想法或原始广告笔记的可选分镜演示素材时。

查看详情添加技能

text-to-3d-creator

用文字描述直接生成3D模型。智能匹配多种3D引擎，可选先生成参考图确认效果再转3D，适合角色、道具、场景等各类3D资产创作。

查看详情添加技能

style-prompt-engineer

分析图片风格或接收风格描述，生成专业风格提示词。支持27种常见风格，中英文自适应，输出可直接复制使用。

查看详情添加技能

social-poster

上传商品图和卖点，生成带专业文字排版的营销海报。自动适配多等平台规格，产品细节零改动。

查看详情添加技能

skill-review

技能的发布前质量检查——结构合规性、描述质量、触发词覆盖、最佳实践审计、评分以及优先修复。

查看详情添加技能

sfx-soundscape-creator

根据文字描述生成环境音、音效、氛围音、ASMR 触发音、游戏音效、影视 Foley 音效等。

查看详情添加技能

scrapbook-photo-illustration

Turn an uploaded photo into a nostalgic scrapbook-style collage: the original photo kept on top, a tiny hand-painted storybook re-creation of the same scene on warm handmade paper below, plus a small handwritten phrase. Image I2I, fixed vertical 3:4. Use when the user wants their photo turned into this "photo-on-top / mini-illustration-on-paper" memorabilia poster style.

查看详情添加技能

realistic-landscape

Realistic landscape photography generator. The user describes a scene in natural language (location, camera position, weather, season, elements, etc.); automatically compiles it into one physically coherent, restraint-leaning, low-saturation photo prompt with natural weather (overcast / cloudy / fog / snow / rain preferred) and calls image-executor to produce the photograph. Style: recorded honestly by a camera, not a postcard or saturated promo shot; only lift color when the user explicitly asks for "vivid / bright". Use when the user describes a landscape and wants a photo / realistic landscape photography. Boundary: single landscape photos only (including vertical / multi-image); not characters, products, illustrations, or other image types.

查看详情添加技能

product-bento-card

将产品照片转化为精美的 8 模块 Bento 网格信息图——液态玻璃磨砂卡片风格，自动提取产品主色调，适用于社交媒体推广帖文。

查看详情添加技能

podcast-production-guide

把任意主题变成播客音频。单人独白可一键生成完整 MP3；多人对谈则按角色分轨输出，方便后期自由剪辑。自动写脚本、确认后合成，中英文均可。

查看详情添加技能

photo-infographic

将照片一键转为精美科普信息图。自动识别主体，提炼关键事实，搭配对称图标与主题色排版。

查看详情添加技能

naive-doodle-poster

Make one poster in the naive hand-drawn doodle family — a single big simplified subject, thick wobbly outlines, flat color on a tinted paper field, hand-lettered text. Ships 8 named preset looks (riso print, quote card, sky-blue home life, grounding PSA, rough-marker monster, fantasy scribble, crayon catalog, ink-music). Use when someone wants a charming, imperfect, hand-made-feeling poster, quote card, fake product-catalog page, or mascot announcement — paste a subject and headline, pick a look, with or without a source image. Also 手绘涂鸦海报 / 儿童蜡笔风海报 / 手写语录卡 / 独立杂志 riso 风海报 / 假产品目录页 / 吉祥物公告 / 照片手绘涂鸦化 / 中文字体海报. Not for photoreal, 3D, or clean corporate vector posters.

查看详情添加技能

megatype-action-poster

Design a big-character action poster — one oversized ultra-condensed headline rendered as a background field that one hero action subject breaks through, as kinetic sports/streetwear/lifestyle campaign key art. Turns a subject, action, prop, and text into a render-ready image_create prompt locked into one original visual system (library of 11 curated looks). Asks which visual system to use when the prompt does not clearly determine one. Trigger when the user wants a sports or streetwear campaign poster, kinetic megatype key art, an oversized-typography action poster, or a big-character poster. Aspect default 16:9; also 9:16, 21:9, 1:1 and other platform-enum ratios.

查看详情添加技能

jumbotron-broadcast

Create a broadcast-frame "caught on the big screen" image that puts one specific face into a live jumbotron / venue IMAG / crowd-cam frame — realistic because it looks degraded and accidental, not polished. Use when the user wants to look like they were caught on a venue big screen (sports scorebug, concert IMAG, news crowd cut, awards audience cutaway) or makes a feed meme ("me on the jumbotron", "react face on the jumbotron", "did you see me in the audience"). Requires a reference face photo. For a polished studio portrait use editorial-fashion-portrait instead.

查看详情添加技能

image-prompt-director

根据参考图、简单描述或一小段剧情，分析并设计具有电影镜头感、概念表达或视觉美感的单幅画面。

查看详情添加技能

ian-xiaohei-illustrations

把中文文章里的判断、流程、状态和隐喻，变成一张张白底、手绘、怪诞但清爽的正文配图。

查看详情添加技能

headshot-portrait-maker

Generate polished single-subject portraits from uploaded selfies and/or a text prompt — LinkedIn / founder / academic / creator-brand / employee-badge. Output count equals the number of reference selfies (one polished portrait per selfie); with only a text prompt, one portrait is generated. Reference images are optional. Triggers on zh or en: "做个 头像", "生成 LinkedIn 照片", "AI 形象照", "把自拍修成职业照", "职业照", "headshot", "LinkedIn photo", "AI portrait", "professional photo", "founder press shot", "profile pic", or "polish this selfie into a profile photo". Boundary: passport/ID photo, group shot, or swap-background-on-existing-photo belong to other flows — see "When NOT to Use".

查看详情添加技能

hallmark

"Anti-AI-slop design skill for greenfield pages, audits, redesigns, and design extraction from URLs or screenshots. Use when the user asks to build a new app or landing page, wants to redesign something, invokes Hallmark by name, or uses audit/redesign/study."

查看详情添加技能

guizang-material-illustration

生成归藏风格的素材插画：带标注的解释型视觉、概念图解、图表美化、数据优先的编辑插图、教育/科学机制图、人文/编辑配图，以及从文章、笔记、截图、产品概念、报告、教学材料或原始数据生成的中文标注素材风格视觉。

查看详情添加技能

frontend-slides

创建零依赖、动画丰富、纯浏览器运行的 HTML 演示文稿。不需要任何安装，打开 HTML 文件就能用。

查看详情添加技能

flat-to-figurine

将扁平插画、线稿、贴纸风格图像转化为立体3D模型。先通过图生图增加体积感和光影，用户确认参考图后，再调用3D引擎生成高质量模型，彻底解决"纸板感"问题。

查看详情添加技能

ecommerce-visual-copywriting

电商主图和详情页的视觉策划技能。根据产品资料、资质边界、目标平台和视觉参考，先判断转化驱动力并确定活动风格锁定，然后出主图/详情页的故事板；用户确认后，提供画面描述、图内文案、设计说明和原图提示。

查看详情添加技能

ec-visual

当用户上传产品图片和视觉参考，用于创建或优化市场主图及产品详情页，同时忠实还原所提供的确切产品及其可见内容时使用。

查看详情添加技能

comic-maker

Turn a story idea into a single multi-panel comic image — manga, American strip, 4-koma, webtoon, or children's-book illustration — generated in ONE image-create call on this platform. Defaults to a 4-panel grid in a single image; the user can request a different panel count in their prompt. Designs the characters, the story beats, the panel order, and any dialogue bubbles all inside one carefully-composed image prompt, then renders the whole comic as a single grid image. Trigger when the user says "make a comic", "draw a comic strip", "generate a manga", "create a 4-panel", "turn this story into a comic", "comic about", or asks for a multi-panel comic image.

查看详情添加技能

cinematic-video-gen

从文本提示或故事脚本生成高质量电影级视频。支持分镜头脚本控制、镜头运动设计、光照效果和场景转场。

查看详情添加技能

chinese-polisher

润色中文文本，修正错字标点、理顺句式、统一术语语气，保留原意与格式，支持轻中重三档力度。

查看详情添加技能

ccd-2000s-snapshot

Convert an image (i2i) or generate a fresh scene (t2i) in the early-2000s CCD point-and-shoot digital camera look — direct on-camera flash, hard shadows, subdued washed-out colors, lifted flat blacks, blown-out soft highlights, slight cool-neutral white balance drift, mild sensor noise (not film grain), soft edges, subtle chromatic fringing, low dynamic range, nostalgic 2000s everyday snapshot. For i2i it preserves the original subject, pose, and composition. Fixed positive + negative prompts — use them verbatim. Triggers: 2000年代CCD卡片机, CCD卡片机质感, 早期数码相机, CCD相机, 2000s digital camera look, CCD point-and-shoot, 直闪硬影, 传感器噪点.

查看详情添加技能

brand-moodboard-grid

Create an editorial brand moodboard — a multi-card grid (default 2×2) of distinct editorial cards unified by a named brand's real identity: exact hex colors, typeface DNA, real slogans/products/campaign facts. Use when the user names a brand and wants a brand moodboard, brand-identity grid, lookbook-style brand spread, Hypebeast/Highsnobiety-style brand feature image, or a brand board for a deck.

查看详情添加技能

brand-kit-video

将品牌的标志、产品和配色转化为一段简短的品牌风格展示视频——同时构建或补全完整的品牌套件。

查看详情添加技能

brainstorming

创意工作前的头脑风暴工具，探索需求、提出方案、呈现设计并生成设计文档

查看详情添加技能

baoyu-cover-image

为文章生成精美封面图，支持5维度定制（类型、配色、渲染、文字、情绪）

查看详情添加技能

art-style-transformation-set

Transform one reference photo into a single multi-style portrait grid: the same face rendered in 4-9 distinct art styles (Ghibli, Pixar, oil, 80s anime, LEGO, Polaroid...) in one frame. Locks the reference photo's facial identity and pose; clothing may change per style. Output is ONE grid image (default 6 styles, 3x2; one result only). For multi-style aesthetic renders, "every version of me" posts, gifts, personal-brand visuals. Reference photo preferred (I2I) for identity lock; falls back to a one-sentence description (T2I).

查看详情添加技能

app-icon-concept-gen

Generate a set of logo designs. The default layout is a 2x2 grid featuring four logo variations; the grid layout adjusts automatically based on the requested quantity. Upon rendering, explain the design concept behind each logo and include a note regarding trademark usage. Selection criteria: Designs should be flat, minimalist, visually focused, and easily recognizable, without incorporating specific text. Trigger commands: "Generate logos," "Create a few logos," "Brand logo," "Logo concepts," "Create a logo based on this website (URL)," "Logo set." Functional scope: This skill generates artistic logo concepts; authorization is required for the reuse of protected trademarks; this skill is not intended for creating UI icon sets or CMYK print output.

查看详情添加技能

ai-travel-snapshot

Put a real face photo into a famous landmark and produce an "I'm on vacation here" AI travel snapshot. Keeps the same person across a batch of 2–4 destinations (#aivacation meme, must be labeled as AI travel). Triggers: "put me in Paris", "send me to XX", "AI travel photo", "fake my vacation", a cute travel pic for a friend, birthday/anniversary "greetings from {a place they'd love}". Boundary: if the user wants a real itinerary/planning guide, use the travel-planner skill (or make that skill). This skill only produces hero visual images.

查看详情添加技能

90s-film-snapshot

Convert an image (i2i) or generate a fresh scene (t2i) in the 1990s Japanese analog film snapshot look — shot on 35mm Fujifilm Superia 400 / Kodak Gold negative. For i2i it preserves the original subject, pose, composition, and framing, changing only the visual style. Fixed positive + negative prompts — use them verbatim, never paraphrase. Triggers: 90s日系胶片, 90年代胶片风, 日系胶片快照, Fujifilm Superia, Kodak Gold, analog film snapshot, 90s film look.