Saying where things go in an image takes a lot of words: a car at the front, a second one behind it, a rock between the trees, branches out of focus in the foreground. The framing ends up taking most of the prompt.
Zones lets you draw it instead. Each box on the frame holds one element of the image, and the prompt tells the story that connects them. In this tutorial, the zones hold text, your own images and snippets, and the same layouts run on several models.
What you will build
- A car chase in a forest, with your own vehicles placed from reference images
- A fashion shoot on a beach, with the character, the sofa and the bag placed from snippets
- A product shot where a thin stream of serum follows the zone you draw
One element per zone
A zone is a box with a name and a short description of what goes there. The prompt calls each zone by its name, like @zone_1 or @sofa, and only has to say what happens in the scene: a car chase on a forest track, a model on a set on the beach. The size of a box sets how big its element is in the frame, so the layout does the work of the framing.
Small or crowded zones are easier to place in Edit zones, which opens the layout at full size.
Images and snippets in zones
A zone does not have to be described in words. Type @ in a zone and pick one of your reference images: it shows inside the zone, so you see which element goes where. In the video, the two vehicles made earlier go into their zones, and the model uses each one in its own area.
Snippets work the same way. A character, an outfit, a product or a color palette from your library goes into a zone, and its card stays visible there. One zone can hold several: the bag in the video is a product snippet with a color snippet for its palette. The snippets tutorial shows how to build your own library.
Native and compatible models
Flux 3.0 and Ideogram 4.5 read zones natively, in their own format, and place each element where you drew it. For the other models, like Nano Banana 2.1, ChatGPT Image or Seedream, Hylope adapts the zones so they follow your layout. These models also take reference images, so they are the ones to pick when a zone holds a photo of your product.
In the video, the same layouts run on native and compatible models without changing the zones.