NanoBanana Trending Prompts
Over 1,400 popular prompts for NanoBanana and GPT Image from X posts ranked by likes, plus a system prompt that turns a short idea into a detailed one
- 746 ★ on the repository
- Last commit
Minimal access
Why this level
- A set of text data and an instruction: it runs nothing and connects to nothing, and the prompts come from other people's posts, so check their rights
Install
- Works in
Manual install
npx meigen init claudeConnects MeiGen MCP, which has this library built in. For Cursor and VS Code use `cursor` or `vscode` instead of `claude`.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- A set of text data and an instruction: it runs nothing and connects to nothing, and the prompts come from other people's posts, so check their rights
This is third-party code. Review the repository files before installing.
What it does
The repository collects popular image-generation prompts from X posts and sorts them by engagement. The file `prompts/prompts.json` holds more than 1,400 entries: rank, identifier, prompt text, post author, likes and views, an image link, the model (NanoBanana or GPT Image), categories, a score, a date and the address of the source post. The same list with image previews and a collapsible prompt text is built into the README, grouped by model. There are six categories: photography, illustration and 3D, product and brand, food and drink, poster design, and UI and graphics. The second part is a system prompt in `prompts/system-prompt-en.md` with a Chinese version beside it. You paste it as the system instruction of a model chat, n8n or GPTs, and it rewrites a short line such as "a bowl of braised pork" into a detailed structured prompt. Its rules: replace feeling words with professional terms, put in measurable parameters like "90mm, f/1.8", state what must not appear in the image, add descriptions of materials, temperature and smell, and structure the text by scene type. The model returns plain text with no markup. The library is also built into MeiGen MCP: through `npx meigen init claude`, `cursor` or `vscode` an agent searches it for generation inspiration for free and without an API key.
Who it is for. For designers, marketers and authors who generate images in NanoBanana or GPT Image and look for proven wording, and for people building their own prompt chains with a system instruction.
Good fit when
- You need prompt examples for product photography, posters or infographics sorted by audience reaction
- A short idea such as "a bowl of braised pork" will not turn into a good prompt and you need a rewriting system instruction
- You want to build your own selection: the `prompts.json` file is easy to filter by model, category and likes
- An agent in Claude Code or Cursor should suggest wording for generation through MeiGen MCP
- You need a sample structure of a long JSON-style prompt for packaging, a poster or technical infographics
Not a fit when
- You need a tool that generates images itself: this is a set of texts, and generation comes from MeiGen MCP or the model
- You need Midjourney prompts with its parameters: the lists cover only NanoBanana and GPT Image
- You need legally vetted templates for advertising: the prompts come from posts by different authors
- An agent has to load the whole file into context: `prompts.json` weighs about 3 MB
Example request
Take the system prompt from the NanoBanana Trending Prompts repository and rewrite my idea with it: a cup of coffee on a windowsill, rain outsideLimitations
- The prompts come from X posts. Each has an author and a source link, but quality and rights to the text stay with the post authors, so check the terms before commercial use. The compiler curates posts by hand: in the cleanup of 9 March 2026 it removed 345 weak ones.
- These are text data, not a tool. The library does not generate images.
- The files are heavy: `prompts.json` is about 3 MB and the README is over 2 MB. Do not load them whole into an agent's context, search by fields or through MeiGen MCP.
- Many prompts contain template fields like `[OBJECT]` or a brand name in Chinese brackets, and you substitute the values yourself. Part of the texts is written in Chinese.
- The CC BY 4.0 license requires attribution, and the license file names MeiGen.ai as the copyright holder. The latest list update is dated April 2026.
How to disable. There is nothing installed here: just stop using the downloaded files and delete the cloned folder. If you connected the library through MeiGen MCP with `npx meigen init`, remove the MeiGen entry from your client's MCP settings.
FAQ
Where do I download NanoBanana prompts as JSON?
In the repository's `prompts/prompts.json` file. Each entry has the fields `rank`, `id`, `prompt`, `author`, `likes`, `views`, `image`, `model`, `categories`, `rating`, `score`, `date` and `source_url`.
How do I turn a short idea into a good image prompt?
Paste the contents of `prompts/system-prompt-en.md` as the system instruction of a model chat and write your idea in one line. The model returns a detailed prompt as plain text without service phrases like "Rewritten prompt".
Which models do these prompts suit?
The list is built for NanoBanana and GPT Image, with several hundred and over a thousand entries per model. By the repository's description the texts also suit Midjourney, but check the result yourself.
How do I search the library from Claude Code or Cursor?
Connect MeiGen MCP with `npx meigen init claude` or `npx meigen init cursor`. It searches this library for free and without an API key, while image generation is a separate function.
Which prompt categories does the collection have?
Six: photography, illustration and 3D, product and brand, food and drink, poster design, and UI and graphics. One entry can belong to several categories.
Can I use the prompts in commercial projects?
The repository license CC BY 4.0 allows use and modification with attribution. The prompts themselves come from authors' X posts, so for commercial work check the terms on the post page through the `source_url` link.
Discussion
No questions or reviews yet.
Related
PromptsNano Banana Pro Prompts by YouMind
awesome-nano-banana-pro-prompts
A large prompt library for Nano Banana Pro: by the author's count more than 10 thousand prompts with examples, sources and a web gallery
Nano Banana Pro Prompts Recommend
AI Image Prompt Recommender
A skill that picks ready image-generation prompts from a large library, shows sample images and assembles a prompt for your text
Awesome Nano Banana Pro
awesome-nanobanana-pro
A short README guide to the best Nano Banana prompts by style: a selection from X, WeChat and Replicate, with no gallery and no templates
A library of 150 GPT Image prompts, each credited to its author with a link to the post, grouped into nine use cases
GPT Image 2 Prompts by YouMind
awesome-gpt-image-2
A prompt library for OpenAI GPT Image 2: more than 2000 curated prompts with previews, updated daily by the author's account
AI Visual Prompt Cookbook
ai-visual-prompt-cookbook
Over 150 ready JSON styles for Nano Banana Pro, GPT Image and Midjourney: copy one style.json and get a consistent look of images
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| NanoBanana Trending Prompts (this entry) | no | 746 | ||
| MeiGen AI Design MCP | yes | API key | 1,793 | an MCP server for generating images and video from an agent: several models, a large prompt library and parallel jobs, while this is only the collection and the system instruction |
| Awesome GPT Image 2.5 Prompts | no | 571 | a library of 150 GPT Image prompts |
README in short
The repository holds an overview in English and Chinese with a description of the collection, a category table and a changelog, the data file `prompts/prompts.json` and two variants of the system prompt. In the overview itself each entry shows its rank, an author with a link, likes, views, an image preview and a collapsible text. The data is maintained by hand, and errors are to be reported through issues. CC BY 4.0 license.