Semantic roles

A role is a short label on a layer that says what it means, not what it looks like. A text layer might be a headline, an eyebrow or a price; a frame might be a product-card or a cta. Roles are optional, but they make every AI feature sharper:

• When you say "make the price bigger", the agent finds the layer with the price role instead of guessing.
• When you apply a brand kit, headline fonts go to headlines and body fonts to body text.
• Resizing and localizing keep related elements together and preserve the visual hierarchy.
• The design description the AI reads groups layers by meaning ("product card: image + title + price + button").

Setting roles#

Select a layer and set Role in the inspector, or ask the agent: "label the layers on this page". When the agent creates layers it always assigns roles and meaningful names. Templates ship with roles on every important layer.

When no role is set, Desyn infers a likely meaning from structure and position (for example, a large text at the top is probably a headline). An explicit role always wins over inference.

Known roles#

These are the roles the AI and the layout tools understand. Any other value is allowed and kept, but treated as opaque.

Roles and template slots#

Templates also mark replaceable content with a slot name (for example headline or photo). Slots are what you fill when you instantiate a template; roles describe meaning. A layer often has both.