← Back to home
Skill

AI Image Prompts · 绘画提示词库

github · 1.0.4 · Downloads 1.1K · Heat 90.0

Source & License

Author: YouMind-OpenLab
License: MIT License (original)
Copyright notice: Copyright (c) 2026 YouMind-OpenLab

This site lists project metadata and a short summary only. We do not republish the author’s original content — follow the source link above for the full text. Verify the LICENSE and NOTICE in the original repository before commercial use.

Show full license text
MIT License

Copyright (c) 2026 YouMind-OpenLab

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

LICENSE · 2026-09-26 23:14:14

Introduction

Recommend curated prompts from a 10,000+ real-world image generation prompt library. Works with ANY AI image model — Nano Banana Pro, Nano Banana 2, Seedream 5.0, GPT Image 1.5, Midjourney, DALL-E 3, Flux, Stable Diffusion, and more. Use this skill when users want to: - Find proven image generation prompts (any model) - Get prompt inspiration for portraits, products, social media, posters, etc. - Create illustrations for articles, videos, podcasts, or marketing content - Browse categorized prompt templates with sample images - Translate and understand prompt techniques

Resource Information

Author:YouMind-OpenLab
Category:Multimedia
Language:en
Version:1.0.4
Auth: No auth
License:MIT
Pricing: Marked free
Publisher:YouMind

License information is collected from official sources and may be outdated or missing. Refer to the LICENSE file in the original repository before commercial use.

Tags
绘画提示词文生图营销素材

Installation

npx skills i YouMind-OpenLab/ai-image-prompts-skill

Features

  • 按用途分成 11 类,不是随机堆砌
  • 全文已入库,站内直接检索、可复制
  • 语义检索(描述需求即可)
  • 内容改写模式(贴正文生成定制提示词)
  • 数据每日同步两次
  • MIT 许可,无需 API Key
Use cases
找某类图片的现成提示词给文章/视频/播客配插画做海报 / 封面 / 社媒图翻译并理解提示词写法

Prompt Library · 22.8K

Source: community prompt library (MIT, synced daily)

15141 distinct prompts across 11 use cases; one prompt may serve several, so category counts add up to more than the deduplicated total.

1099 matches

Isometric 3D Cube Prompt

others

A user mentions finding and successfully using an AI prompt for generating an isometric 3D cube in Nano Banana, indicating a simple visual generation task.

Show full prompt
an AI prompt for an isometric 3D cube

Photorealistic Leaf Shaped Like a Lion

others

A simple, direct prompt for generating a photorealistic image of a leaf that is shaped like a lion, focusing on the fusion of natural elements and animal form.

Show full prompt
A photorealistic image of a leaf that looks like it’s in the shape of a lion.

Structured Prompting Guide for Nano Banana Pro

others

A guide on how to structure prompts for Nano Banana Pro to achieve better results, emphasizing the use of specific roles, concrete descriptions (e.g., lighting), and structured formatting like bullet points or mini sections.

Show full prompt
• tell it the role

“you are a senior art director / product photographer / magazine designer”

• be concrete, not poetic

bad: “make it cinematic”
good: “soft studio lighting, shallow depth of field, editorial look”

• lock the format

image type, layout, aspect ratio, style
don’t let it guess

• describe what matters, not everything

subject
background
lighting
style
that’s it

• if you want accuracy → structure

bullet points or mini sections work better than paragraphs

Photorealistic Leaf Shaped Like a Dragon

others

A simple, photorealistic prompt for generating an image of a single leaf manipulated or naturally shaped to resemble a dragon.

Show full prompt
A photorealistic image of a leaf that looks like it’s in the shape of a dragon.

Surreal Horror Portrait Revealing Skull

others

A highly detailed prompt for generating a hyper-realistic, cinematic concept art portrait in the surreal horror aesthetic. The image depicts a young woman pulling her facial skin apart to reveal a realistic human skull underneath, emphasizing visceral texture details, sweat droplets, and dramatic, moody lighting.

Show full prompt
{
  "prompt": "Hyper-realistic, cinematic concept art portrait of a young woman with dark hair pulling the skin of her face apart with both hands, revealing a realistic human skull underneath on the right side of her face. The exposed skin appears stretchy, wet, and visceral, with visible sweat droplets, lifelike skin pores, and extreme texture detail. She has a wide-eyed, intense expression conveying shock and horror. Dramatic, moody cinematic lighting with strong contrast, deep shadows, and a dark, minimal background. Surreal horror aesthetic, macro photography look, ultra-detailed, 8K resolution, photorealistic finish.",
  "negative_prompt": "cartoon, anime, illustration, 3d render, CGI, drawing, painting, blurry, low resolution, low quality, bad anatomy, distorted fingers, extra fingers, deformed hands, mutated face, text, watermark, logo, signature, messy composition",
  "aspect_ratio": "3:4",
  "style": "surreal horror, cinematic, hyper-realistic",
  "quality": "ultra high",
  "resolution": "8k",
  "signature": "{argument name="signature" default="Shreya Yadav"}"
}

Photorealistic Leaf Shaped Like a Bird

others

A simple, photorealistic prompt for generating an image of a single leaf manipulated or naturally shaped to resemble a bird in flight.

Show full prompt
A photorealistic image of a leaf that looks like it’s in the shape of a bird in flight.

Photorealistic Miniature Woman on Milkshake Cup

others

A photorealistic image generation prompt for a surreal scene featuring a miniature woman sitting casually on the lid of a colossal strawberry milkshake cup on outdoor pavement. The prompt emphasizes playful posing, pastel clothing, and oversized props like a handbag and sunglasses.

Show full prompt
a photorealistic scene of a woman sitting casually on the lid of a large strawberry milkshake cup placed on outdoor pavement. Keep the pose relaxed and playful. Dress her in a pink fitted top, a white skirt, and matching pastel shoes. Add an colossal glossy pastel-pink handbag with a gold chain strap placed near the cup, much larger than the tiny woman. Place large heart-shaped pink sunglasses on the ground beside her, also oversized compared to her. The milkshake inside the clear cup should look thick and bright pink. Soft natural daylight

Scandinavian Folk-Art Depiction

others

A simple image generation prompt to create an illustration in the style of Scandinavian folk art, characterized by minimal shapes, soft pastel colors, stylized florals, and a woodblock-print texture with a storybook feel.

Show full prompt
A Scandinavian folk-art depiction of {argument name="subject" default="[SUBJECT]"}, using minimal shapes, soft pastel hues, stylized florals, and cozy symmetrical motifs. Woodblock-print texture with a charming, storybook feel.

4K Upscale and Detail Enhancement Prompt

others

A prompt designed for high-quality upscaling, instructing the model to convert an uploaded image to 4K resolution while strictly preserving 100% character consistency, facial features, clothing, composition, and original lighting. It specifically demands enhancement of fine details like skin texture, hair strands, and fabric texture with sharp but natural edges.

Show full prompt
Upscale the uploaded image to 4K resolution, preserving 100 % character consistency, original facial features, clothing details, and scene composition. Maintain original lighting, texture, and fine details such as hair strands, fabric texture, reflections, and shadows. Apply high-quality upscaling with sharp but natural edges, no distortions, and minimal AI artifacts. Keep the character’s identity, proportions, and expressions unchanged. Enhance clarity in skin texture, eyes, and background elements while preserving the original color palette and artistic style. Output a crisp, highly detailed, ultra-high-resolution image suitable for print and professional display.

Cinematic Night City with Visible Power Grid

others

A prompt for generating an ultra-real cinematic night view of a modern city where the normally invisible electrical power grid is faintly visible, subtly connecting the buildings and streets under realistic lighting conditions, aiming for grounded film realism.

Show full prompt
Ultra-real cinematic night view of a modern city where the invisible electrical power grid becomes faintly visible in the air, subtly connecting buildings and streets under realistic lighting and grounded film realism.

Spaghettification Glitch Effect (Loki Inspired)

others

A multi-layered prompt designed to recreate the 'spaghettification' or disintegration effect seen in the Loki TV series. It uses four distinct layers of instructions to ensure background stability, violent unraveling of the subject's edges into non-glowing ribbons, analog glitch overlay concentrated only on the edges, and ghostly motion-blur afterimages behind the subject.

Show full prompt
(Photorealistic:1.3), raw photograph, cinematic lighting, sharp focus on the subject’s face and eyes, crisp skin texture.
LAYER 1: BACKGROUND STABILITY (CRITICAL):

The entire background environment from the input image remains completely STATIC, FROZEN, CLEAN, and UNCHANGED — no noise, no scanlines, no distortions, no glitches in the background.
LAYER 2: FOREGROUND SUBJECT EDGE-CHAOS:

The main subject’s core (face/torso) stays clear and undistorted, BUT the outer silhouette boundaries are unraveling violently into long ribbons of (non-glowing physical matter:1.3), showing stretched fabric/leather/latex-like material with realistic strand separation and tearing flow (“Spaghettification”), no gore.
LAYER 3: ANALOG GLITCH OVERLAY (EDGE-ONLY):

Heavy VHS static, scan lines, color bleed, and screen tearing noise are CONCENTRATED EXCLUSIVELY ON THE UNRAVELING EDGES of the main subject, like an edge mask. No full-frame VHS overlay; the core and background remain clean.
LAYER 4: GHOSTLY AFTERIMAGES (BEHIND SUBJECT):

Directly BEHIND the unraveling main subject, visible through the gaps, there are TWO translucent, ghostly motion-blur AFTERIMAGES. They are violently twisted into agonizing, struggling postures, dragged by temporal distortion. They remain behind the subject, partially occluded, not separate characters.

Negative prompt (optional but recommended):

full-frame glitch, background distortion, extra people, duplicate main subject, neon glow, energy beams, particles, blood, gore, exposed organs, messy face, blurry eyes, deformed hands

Show Interior of Building

others

A simple prompt instructing the AI to generate the interior view of a building based on an uploaded image of its exterior.

Show full prompt
Show me the interior of this building @img1
Prev 64 / 92 Next

Prompts are model-agnostic and work with Midjourney / DALL-E / Flux / Stable Diffusion.

Dependencies

- Node.js 18+
- 无需任何 API Key(数据全部公开)
- 提示词已完整同步到本地与数据库,站内检索不依赖第三方

Supported platforms

claude-codecursorcodexgemini-cliopenclaw

Changelog

Watching official sources
Content update 2026-10-21

内容有更新(细节请以官方文档为准)

Content update 2026-10-20

内容有更新(细节请以官方文档为准)

Content update 2026-10-19

官方源有内容更新(首次纳入监控)

Information boundary

Information on this page is collected from official documents for technical presentation only and does not imply our endorsement. Pricing, quotas and availability are subject to the latest official statements. We make no commitment regarding any action you take after leaving this site.

Similar skills

视频短视频图片音乐音频配音口播数字人全能创作

Marked paid

通过同一个AI创作工具完成AI图片、视频、音乐与语音创作,查找公开社交数据,编辑视觉成果,并集中查看和管理生成资产。Beatra 全能创作支持AI图片生成(文生图)、AI视频生成(图生视频)和AI视频编辑,通过AI音乐生成把想法做成歌曲或纯音乐,也可用AI语音生成(文字转语音)、多语言旁白和AI语音克隆完成跨媒介内容;还能查找抖音数据、小红书数据,以及 YouTube、TikTok、Instagram 与 X 上的公开帖子、笔记、评论、账号和趋势,把社交媒体数据用于竞品调研和创作。它是一款面向AI内容创作与多模态内容创作的AI创作套件,用起来就是一个把图片、视频、音乐与语音收在同一处的AI创作助手,适用于社交内容、产品展示、营销素材、课程、播客、短片和跨媒介项目;还能在不同媒介间复用原始素材、已有成果和查到的公开资料,查看制作进度、返回文件、实际用量和积分记录。

AI数字人口播与虚拟主播视频

Marked paid

用一张人像和一段短脚本或语音,制作有明确讲述内容的数字人口播视频。可使用现成录音,也可以先选择音色,把脚本生成旁白,让画面中的人物随声音完成稳定、自然的口播。适合 AI 主播、虚拟主播视频、数字人讲解、让照片说话、照片说话视频、产品讲解视频、数字人培训视频、公告和社交口播,并在交付后检查人物一致性、声音清晰度、口型同步和动作稳定性。

架构图一键生成

Marked paidAPI key required

使用 ContextWeave 生成或修改架构图、流程图、思维导图和复杂信息图。适用于需要从代码、文件或自然语言中提取结构与关系并实际产出 CW、SVG、HTML、可编辑 PPTX 或原生 VSDX 的请求;未特别说明时,PPTX 与 Visio 使用原生形状和连接器导出。不适用于统计图表、手绘插画或像素级排版。

罗斯蒙德可视化图形生成

Not specified

【罗斯蒙德可视化图形生成】南京罗斯蒙德基于已交付项目沉淀的可视化图形生成场景:根据业务描述或系统需求,自动生成可编辑的架构图、流程图、ER 图、时序图、状态图与 C4 图。采用「统一图模型」一次建模、多格式输出:结构化文本(Mermaid / PlantUML,可嵌入 Markdown、进 Git 版本管理)、可编辑图形文件(drawio / Excalidraw,可拖拽修改节点与样式)、静态图(SVG 自渲染,PNG 给转换命令),并输出图形说明(图的含义 + 关键节点解释)。内置样式系统统一配色与命名,内置布局引擎自动排布,校验脚本拦截孤立节点、悬挂连线与语义缺漏。触发词:画架构图、画流程图、画 ER 图、画时序图、画状态图、C4 图、系统架构图、部署架构图、数据流程图、实体关系图、生成 drawio、生成 Excalidraw、导出 mermaid、导出 PlantUML、导出 SVG、导出 PNG、可编辑的图、图转可编辑、架构图可编辑、流程图二次修改、画图、出图、画个图。

海报设计工作室

FreemiumAPI key required

将一段活动描述、一张产品照片或品牌参考图,转化为抓人眼球的活动海报、宣传 banner 或社交媒体配图。这款 AI海报设计与制作工具生成图片丰富的海报,搭配清晰视觉层级、按品类匹配的风格和预留文字区域,覆盖活动、音乐节、电影、产品发布和促销场景。你可以把它当作 AI宣传单制作工具来制作促销宣传单,当作活动海报设计工具,当作电影海报生成器,当作音乐节海报设计工具,当作产品发布会海报的 banner设计 AI,或当作平面设计海报工作室。从活动主题出发,用品牌风格参考图变换产品或场景照片,或在已认可草稿上做定向精修,完成可印刷或可分享的成品。

架构图一键生成

Marked paidAPI key required

使用 ContextWeave 生成或修改架构图、流程图、思维导图和复杂信息图。适用于需要从代码、文件或自然语言中提取结构与关系并实际产出 CW、SVG、HTML、可编辑 PPTX 或原生 VSDX 的请求;未特别说明时,PPTX 与 Visio 使用原生形状和连接器导出。不适用于统计图表、手绘插画或像素级排版。

What is a Skill

A Skill is a packaged single capability, such as text recognition, document summarization, or keyword extraction. Developers wrap the calling logic and prompts into a plugin-like unit, so users can load it without understanding the underlying model. StackPilot indexes a large number of skills that can be tried instantly or assembled into a plan.