From viral images to prompts you can actually reuse
Prompts we have already run
Every prompt here was executed before we published it. The image you see is what came out — not one somebody uploaded without checking the prompt still works.


- 461
- verified cases
- 461
- images of our own
- 397
- with editable variables
- 39
- before/after effects
Differences
What the others do not have
Run, not collected
All 463 prompts were fired at the model before being published. The image on each entry is what that exact prompt produced, not something uploaded by someone who never checked it still works.
A check that actually blocks
Every image goes through a vision check that reads the text inside it. If lettering shows up in another script, or an undeclared third-party brand appears, the entry does not publish. Three are held back today.
Prompts with variables, not dead text
City, language, colour, how the person looks: every slot is a typed field with examples. Change what you like and the prompt rewrites itself in front of you, ready to copy.
Spanish written, not machine-translated
Titles, descriptions and examples written in Spanish rather than run through a translator from Chinese. Same for English. No other library of this kind bothers.
Our own images
462 of 463 images were generated and paid for by us. None come from someone else's gallery, so there is no rights claim waiting for the day this grows.
Open data
The prompts and their taxonomy live in a public MIT-licensed repository. Clone them, build your own tool on them, or correct us with a pull request.
How it works
From template to finished prompt, in three steps
1Pick a template
Every case is verified and sorted into ten categories. Each entry states which model tier it needs and whether moderation refuses it.
2Fill the variables
Prompts come parameterised. Change city, brand, language or how the person looks with one click, and the text rewrites in front of you.
3Copy and generate
You take the resolved prompt, with no brackets left for the generator to draw literally.
Effects
Effects that start from a photo of yours
39 templates that transform an image instead of creating one from scratch. Drag the divider to compare before and after.
From the catalogue
A selection
Products & CommerceCopy promptPremium Meat Seafood Hero
12 varsoptimizado
Photography & RealismCopy promptPastel Flower Street Portrait
6 varsfree
Data & DocumentsCopy promptAncient Philosophers Timeline Chart
4 varsfree
Brand & LogosCopy promptMatcha Brand Identity System
3 varsoptimizado
UI & InterfacesCopy promptUrban Metabolism Infographic Atlas
2 varsoptimizado
Posters & TypographyCopy promptPremium Wildlife Science Poster
14 varsoptimizado
Illustration & ArtCopy promptFantasy Character Portrait Grid
10 varsoptimizado
Characters & PeopleCopy promptSunny Youth Pixar-Style Portrait
7 varsfree
What it costs
The whole catalogue is free
And it will stay that way. Charging to read a prompt breaks the one thing that sets us apart, which is showing you it works. If we ever charge, it will be for generating in here, without the copy and paste.
Catalogue
$0forever
- All 463 prompts in full
- Variable editor and shareable link
- Before/after effects
- English and Spanish
- No account, no sign-up
Generate here
PlannedUndecidednot yet
- Generate straight from a filled-in entry
- Keep your results
- Bring your own API key, at no cost
- Publish what you make to the catalogue
No waiting list and no date. When it exists, it will say so here.
Questions
What people usually ask
Is it really all free?
Yes. All 463 prompts are complete on their entry, copy with one button, and need no account. There are no hidden premium prompts and no free-copy counter.
What does a verified prompt mean?
That we ran it against gpt-image-2 and the image on the entry is what that exact prompt produced. It also passes an automatic check that reads the text inside the image and spots third-party brands; whatever fails does not get published.
Do they work in Midjourney, Nano Banana or Stable Diffusion?
The prompt is English text, so it pastes anywhere. What we guarantee is only what we tested: gpt-image-2. Each entry states the model tier it needs, because cheaper models do not render typography reliably.
Can I use the images in a commercial project?
The images are ours and we publish them as evidence of what the prompt does. What you take away is the prompt: generate your own image and that one is yours. Watch the 74 entries flagged with a third-party brand, which carry someone else's logo inside.
Where do the prompts come from?
From open repositories under MIT and CC0, and from creators who published them on X and elsewhere. Every entry links to whoever wrote it. We rewrite them into English, parameterise them and run them.
Can I contribute mine?
Not yet, but it is planned. Meanwhile the repository is public: a pull request with your prompt is the fastest route and the credit stays yours.
Open source
The prompts are yours too
The catalogue and its taxonomy are on GitHub under MIT: prompts in JSON, typed variables, categories and the editorial policy we follow. Clone it, use it in your own tool, or correct us.
View the repository- 461
- prompts in JSON
- MIT
- MIT licence
- 2
- languages







