Pictures and vector

The design agent can make pictures with an image model and put them on the canvas, or draw artwork as vector layers whose parts can move on their own. You set it up under Images in the agent's settings; pictures use your own credit with the image service.

Step by step

1. Open the agent's settings

   Open the Agent panel (Agent in the top bar) and click the cog at its top. The agent needs to be connected with an API key first: see Connect your own API key.

2. Pick an image model

   Under Images, Provider is the service that makes the pictures: the agent's own when it makes them (OpenRouter, OpenAI or Gemini), or another one with a key of its own. Pick a model in Image. With None, the agent draws everything as vector shapes.

3. Test it

   Click Test in Images: it makes one small test image, which uses a little of your credit. With Transparent background on, it checks that too: when a model can't make one, SVGcinema asks it for a flat green background and removes that.

4. Ask for a picture

   Back in the chat (‹), ask for a picture. Before making one, the agent shows a card with the prompt (you can edit it), the model, the size, the background and about what it costs, when the price is known.

5. Generate it

   Click Generate, or Cancel. The picture goes into the project's Assets and the agent puts it on the canvas: one layer, which moves as one piece. Don't ask again skips the card; Ask first in Settings brings it back.

6. Vector for parts that move

   With Prefer vector when parts move on, under Images, the agent makes artwork whose parts move on their own as vector: separate, editable layers. Vector says who draws it: the agent itself, or a vector model (on OpenRouter or a Custom service).

Keys