grok-imagine-video is the video generation entry point for the xAI Grok Imagine series, suitable for turning text ideas or existing images into dynamic clips. It focuses on text-to-video and image-to-video: the former starts with a scene description, while the latter uses an image as a visual starting point and adds motion intent. Through this platform, you can submit generation tasks, track their status, and obtain video links, making it suitable for short-video assets and creative previsualization.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
api.acedata.cloudgrok-imagine-videoInput parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
The creation method and task management descriptions correspond to this platform's API endpoint; aspect ratio, duration, and resolution settings are not equivalent to the model's native limits.
When no image is available, you can use prompts to describe the subject, environment, action, and camera intent to start text-to-video creation. This is suitable for first exploring whether a visual concept works before deciding whether to move into filming or post-production. Prompts should focus on clear actions and avoid packing multiple unrelated scenes into the same description.
When you have product images, character illustrations, or scene images, you can submit an image through image_url and then describe the desired action in text, such as the subject turning around, the camera slowly pushing in, or changes occurring in the environment. The image provides a visual starting point, while the prompt adds motion direction, making it suitable for animating existing design assets.
Generation tasks can be submitted asynchronously, without requiring business pages to continuously wait for complete results. After saving the task_id, an application can query progress or use a completion callback to receive results, then display the video link based on the status. This workflow is suitable for asset workbenches, creative review pages, and automated content production programs.
Enter a shot concept, specifying the subject, action, and atmosphere in the frame, to generate a video draft for discussion. Teams can further refine options around motion direction, compositional intent, and pacing choices. The deliverable is dynamic previsualization material, not a fully finished video with subtitles, music, and editing already completed.
Use existing product images as input, add descriptions of display actions or camera changes, and create candidate dynamic assets. Suitable for exploring ways to present product displays, campaign visuals, or brand content. Before formal release, check each item to ensure product outlines, logos, and details still meet the original design requirements.
Submit character illustrations or environment images, and use action prompts to describe changes in facial expressions, poses, or scene movement, obtaining clips for creative discussion. Adjust descriptions around the same input, compare different motion approaches, then select suitable results for subsequent editing, rather than expecting complex narratives to be completed in one pass.
If a project includes both text-only concepts and existing images that need animation, the two creation modes of grok-imagine-video make it easier to organize a unified workflow. Compared with grok-imagine-video-1.5:official, the latter explicitly requires image input and is not suitable as a direct replacement for text-only ideation. When choosing, first determine the starting point of your assets rather than looking only at the version name.
grok-imagine-video, public invocation IDs with :official or :reverse, and variants containing 1.5 should be managed as separate configurations and should not be treated as completely interchangeable names. When reusing an existing workflow, explicitly specify model; when migrating to a related version, retest input modes and parameter combinations, rather than directly applying new-version settings to this model.
For text-to-video, provide a prompt; for image-to-video, use image_url to define the visual starting point and add action descriptions. Configure reference image capabilities according to the guide for this public variant.
Specify model=grok-imagine-video for /grok/videos, starting with a small 6-second task; use the current parameter range in the API panel for duration, and do not directly apply duration specifications from other variants. Choose aspect_ratio and resolution according to the actual visuals.
Use async or callback_url to track the task, and use task_id to query /grok/tasks; wait for succeeded before reading video_url, check the subject, motion, visual continuity, and ending, and proceed to editing with the video actually returned.
A wooden toy train moves slowly along a tabletop, with the camera following at a low angle, maintaining the train's appearance and a simple background.
Explicitly enter the complete ID without a suffix, and set parameters according to the API panel; do not directly treat the duration or resolution of another 1.5 variant as guaranteed for this ID.
You can start creating text-to-video content from a text prompt. Simply describe the subject, environment, and actions you want to appear; there is no need to prepare a static image first. If you already have a clear visual design, you can instead use image input and focus the prompt on motion and camera changes.
Prioritize describing what happens next in the image, such as a person turning around, an object moving, or the camera pushing in, rather than merely repeating what is already in the image. After submitting an image through image_url, action prompts can help convey creative intent; the result should still be checked for subject appearance and visual details.
They are different invocation IDs. grok-imagine-video is intended for creative workflows starting from text and images, while grok-imagine-video-1.5:official explicitly uses image-to-video and requires image_url to be submitted. When migrating, choose again based on the input material rather than simply replacing the name.
Do not submit these names as dedicated switches for this model. When creating, naturally describe the desired atmosphere, actions, and presentation in the prompt, such as relaxed, restrained, or dramatic; style descriptions express generation intent and are not deterministic effect presets.
Save the task_id returned by the submission, track the status through task queries, or set callback_url to receive completion notifications. When the state in the result is succeeded and video_url exists, then display or retrieve the video; pending means it is still being processed, while failed should enter the error-handling flow.
Choose a development approach or application suited to this model and follow its setup guide.
Python SDK Integration Guide
SDK installation, authentication and examplesTypeScript SDK Integration Guide
SDK installation, authentication and examples