Video Generation

To generate AI videos from text prompts using z-ai-web-dev-sdk, you create an asynchronous task with your natural language prompt, then poll the task with exponential backoff until the video generation completes and results are retrieved.

Yes, you can create videos from base64-encoded images. The image-to-video workflow supports animating single base64-encoded images or generating smooth transitions between start and end keyframes.

Asynchronous video generation works by creating a task, polling the API with exponential backoff to check status, and retrieving the finished video output once the long-running task completes successfully.

Text-to-video generation supports configurable resolution, frame rate (fps), duration, and quality modes. You can specify these parameters when creating the asynchronous task via the backend SDK.

Yes, a backend environment is required. The video generation SDK is designed for backend-only usage to handle asynchronous task creation, polling, and result retrieval without exposing API keys or processes to the client.

Asynchronous video generation is limited by long-running task times requiring polling, basic error handling, and a focus on short-form content creation, making it unsuitable for real-time video streaming or long-duration video outputs.