Describe it in words, get a picture

“Draw a cat” gives you a random result: the artist invents the style, the light and the composition differently every time. A model sits between the request and the artist, and that buys more consistency than switching artists does.

What is inside

1. What to draw — the entry, an ordinary request in your own words.
2. A brief for the artist — the model. It expands the request into a detailed description: subject, style, lighting, composition, mood.
3. The picture is drawn from that brief.
4. The postcard — the exit, with the file attached.

The brief is written in English. Image models are trained largely on English captions and read them more precisely; you still write to the bot however you like.

What to set up for yourself

• The first step’s prompt is your style. Add “in watercolour”, “flat illustration”, “no text in the image” and every postcard starts to look like a set, which is usually the point.
• The image model is named explicitly: an empty field on a media node means “no model selected”.
• Size and other parameters appear in the form once a model is picked: they differ between artists.

What to connect

Nothing. Drawing is paid for in quarks from the balance; no separate key. One picture costs noticeably more than an ordinary answer, so a bot like this left open to the public can surprise you with a bill faster than you expect.