Meme Generator

Because they are photographs, they belong to whoever took them, and most of the famous ones are pictures of real people who never agreed to be a punchline for a decade. What is not anybody's property is the rhetorical shape — reject one thing and prefer another, compare expectation with reality, label the parts of a scene. This app builds those shapes out of new artwork. Name a template and it says so and offers to rebuild the shape in one click.

Because image models cannot reliably spell, cannot be told to hit a character limit, and cannot be edited. The art brief explicitly forbids words, letters and lettering inside the picture; the caption is then typeset onto it on a canvas in your browser. That makes the words correct, legible at thumbnail size, identical in everyone's browser, and free to change as many times as you want after the picture is paid for.

Five things, all refused in the browser before any credit is reserved: pile-ons and the machinery of aiming a crowd at somebody; contempt for a group of people for what they are; a meme aimed at a real, named individual; sexual or suggestive content; and anything sexual, suggestive or hostile involving a child. A refusal quotes the exact words that caused it.

On the fourth of those the app deliberately does not try to sort public figures from private ones — it cannot do that reliably in a browser, and the cost of getting it wrong falls on the person with no platform.

Two limits, stated rather than implied. The guard is structural rather than a word list — it reads the shape of a request rather than matching banned words — so it catches the clear cases and coded or implicit versions can get past it. And it runs in the page: anyone driving the API directly is not behind it, and only the model's own refusals apply there. The app says both of those plainly rather than implying total coverage.

Yes, for a real part of it. Writing your own captions, choosing a layout, typesetting them over a browser-drawn plate, the lint, the content guard, the full art brief and the PNG export all run locally and cost nothing. The two metered runs are the caption writer and the picture. The bundled example replays a real writing run for free.

The app quotes it at the button before anything is spent, reading the figures from the platform rather than repeating a number from documentation. Two runs are involved: a small text run for the captions and a per-image run for the picture. The figure shown is the reserve — the amount held while the run happens — and the settled charge varies from picture to picture against an identical reserve, which is why the app quotes the reserve and never a price.

As often as you like, for nothing. The picture and the words are separate all the way through: re-typesetting happens on the canvas, so switching to one of the alternative versions, rewriting a line, moving a label to another corner or turning capitals off all redraw instantly with no run. Memes saved to your account keep their words, so a meme reopened next month can be re-captioned without repainting it, and past memes are searchable by meaning rather than by keyword.