What these tools actually do

An image generator turns a description into a picture by starting from noise and removing it, step by step, towards something that matches your words. It is not searching a library and it is not copying one photograph. That is why you can ask for a rickshaw on the surface of the moon and get one.

What follows from that is worth knowing before you start.

It has no idea what things are. It has learned what they usually look like. Hands, text, reflections and the number of legs on a chair are all places where "usually" is not good enough. Newer models are better at hands. Almost all of them are still poor at text inside the image.

The same prompt gives a different picture every time. Unless you fix the seed, which we cover in lesson three, you are sampling, not retrieving.

It is confidently average. Ask for "a beautiful logo" and you get the middle of everything it has seen, which is exactly the look people now recognise instantly as machine made. Specificity is the entire craft.

The practical use in a working week is narrower than the demos suggest, and more useful. Backgrounds and textures. Concept images for a pitch, so a client can react to something instead of nothing. Social posts where the picture supports the words. Mockups of a product idea before anyone builds it.

Where it is still weak: anything needing accurate text, a consistent human face across many images, or a real place shown truthfully.