LemGen Trending Prompts

A clean, source-linked prompt dataset for AI image/video agents, MCP servers, and creative products.
_{5,000 curated prompts · GPT Image · Seedance · Nano Banana · Midjourney · image/video previews · LemGen source URLs}

Dataset · Schema · Usage · MCP Server · 中文

---

What Is This?

This repository is the public data layer behind lemgen-ai-design-mcp.

It provides an agent-friendly JSON dataset of AI image and video prompts. Each entry keeps enough context for real creative tooling: title, prompt, model, tags, preview media, engagement fields, ranking metadata, and a source URL back to LemGen.

Use it for:

• prompt search in MCP servers;
• AI image/video inspiration panels;
• prompt engineering examples and benchmarks;
• model comparison datasets;
• creative agent memory;
• prompt packs, galleries, and SEO content pages.

Prompt Packs

These hand-edited packs turn the larger dataset into narrow, reusable content assets for creators and ecommerce teams:

• Perfume product photography prompts

Preview

Dataset

data/trending-prompts.json

Current snapshot:

Type	Count
Image prompts	3,264
Video prompts	1,736
Total	5,000

Model coverage:

Model	Count
GPT Image	2,710
Seedance 2.0	1,736
Nano Banana Pro	409
Midjourney	80
Other image models	65

Schema

type GalleryPrompt = {
  id: string;
  rank: number;
  title: string;
  prompt: string;
  author: string;
  author_name: string;
  handle?: string;
  likes: number;
  views: number;
  image: string;
  images: string[];
  video_url?: string;
  media_type: "image" | "video";
  model: string;
  categories: string[];
  rating: number;
  score: number;
  date: string;
  source_url: string;
};

Usage

Import the dataset:

import prompts from './data/trending-prompts.json' assert { type: 'json' };

Find product photography prompts:

const productPrompts = prompts.filter((item) =>
  item.categories.some((name) => /product|brand|commercial/i.test(name))
);

console.log(productPrompts[0].prompt);

Search by natural language keywords:

const query = 'cinematic perfume product photography';

const results = prompts.filter((item) =>
  [
    item.title,
    item.prompt,
    item.model,
    ...(item.categories || []),
  ]
    .join(' ')
    .toLowerCase()
    .includes(query.toLowerCase())
);

Split image and video prompts:

const imagePrompts = prompts.filter((item) => item.media_type === 'image');
const videoPrompts = prompts.filter((item) => item.media_type === 'video');

Companion MCP Server

Use this dataset directly inside AI coding tools:

https://github.com/aithink001/lemgen-ai-design-mcp

The MCP server adds search, full inspiration lookup, model guidance, multilingual prompt tools, local preferences, and optional LemGen generation.

Refresh

This dataset is exported from the LemGen application database:

cd ../lemgen
pnpm lemgen:mcp:maintain

The export pipeline keeps public URLs on https://lemgen.org, verifies JSON shape, and runs a brand-clean scan before publishing.

License

MIT