This one surprised me when I tested it.
The Veo 3.1 API accepts both a start image and an end image. You give it the first frame, the last frame, and a prompt and it generates the video in between. Built a workflow with a simple form that handles both optional image uploads, converts them to base64, and passes them through correctly.
The output quality when you anchor both ends of the video is noticeably more controlled than pure prompt generation alone.
Has anyone tested image to video generation seriously yet? Would love to know what use cases you are finding for it.