About this page

The words go to the library, which picks the model and the settings. What comes back is a photograph like any other: private until you say otherwise, carrying the model, the words and the seed that drew it, and the permanent tag Made by AI. Drawn from a photograph of your own, it is as visible as that photograph, and goes into its galleries straight after it.

Working from a photograph: put one in hand (from your library, from the public photos, by uploading it or by pasting it) and the words become what to change about it rather than what to draw. It comes back the same shape as your photograph, exactly, and as big as the card can draw in a handful of seconds: about 1.2 megapixels, trimmed by a per cent or two at the edges to land on the shape rather than stretched or shrunk to reach it. Qwen-Image 2.1, Klein and LongCat (one photograph at a time) can be shown an image; the others draw from words alone and are set aside while one is in hand. Removing the background first is one button, and the cut-out is a new photograph you can then draw from.

Licensing: Klein, Z-Image and LongCat are Apache 2.0, with nothing to think about. The FLUX.1 dev, FLUX.2 dev and Klein 9B styles are under Black Forest Labs' non-commercial license, and Qwen-Image 2.1 under Alibaba's research license, which says the same thing in different words. The line falls in an unobvious place in both: the images may be used for anything, commercially included, but drawing with the model for a business needs a license from whoever made it. Removing a background uses BiRefNet, which is MIT and unrestricted.

What this is for: original art, decoration and brand art. Never an image of a game on a viki, and never an image of a real person or event on Heartipedia. Those pages show records of something; a drawing of one is a image of a thing that never happened, in the place a reader expects the thing itself.