Why a generation takes a while

What you see while it runs

Click Generate and a tile appears in the generation tray with a spinner and a timer counting up (12s, 1m 5s). There is no estimate, because the time depends on the model, the number of reference images and how busy the provider is. When the image is ready the tile shows it with Click to place; nothing lands on the board until you place it. A failed one shows Failed and a toast that explains why.

What affects the time

• The model. Each entry in the model menu is a different provider and speed; if one is slow today, try another for the same prompt.
• Reference images. Every @mentioned element, attached style and uploaded image is sent along with the prompt, and more references mean a longer request.
• No background. The cutout step runs after the image is made, so transparent results take a little longer.
• Demand. When a provider is overloaded the request fails with Service unavailable due to high demand rather than waiting forever; switch model or retry in a minute.

Working while you wait

Up to five generations can run at the same time; a sixth is refused with Up to 5 generations at a time. Wait for one to finish. Fire off a batch, keep editing, and place the ones you like as they land. The tray lives on the canvas view for this session; leave the project and unplaced results are gone, though every generation is kept in your AI history. The app waits up to ten minutes for a single result before it gives up and reports a failure.