Collect generations into visual pipelines: prompt, image, video, variations, and comments stay on one canvas.
Create new canvas
