Image-to-video is asynchronous: submit a task to get a task_id, then poll until it finishes and returns the video URL. The two steps below use the same endpoint and parameters as the usage docs page.
Video tasks finish asynchronously (about 60-90s, up to 4 minutes at peak); the image in the example must be replaced with Base64 or a data URL, and the API key is a placeholder.
Upstream protocol references: OpenAI API reference · Anthropic (Claude) API overview · Google Gemini API docs
The response returns a task_id used for polling in the next step.
Poll until the status becomes complete; the response then carries the video URL.