grok-imagine-video:reverse is the video generation entry point in the xAI Grok Imagine series, supporting video creation from text descriptions as well as dynamic shot design starting from static images. It is suitable for producing 1–15 second creative clips, product showcases, and storyboard previews. When creating, organize your input around the subject, action, scene, and camera movement; after generation, obtain a video link to integrate into editing or content publishing workflows.
Clarify capacity, inputs and outputs, and invocation methods before choosing a model.
The durations and operating methods above apply to this API entry point and do not treat the native specifications of other versions in the series as guarantees for this model.
Learn what grok-imagine-video:reverse can bring to your work.
When no ready-made image is available, you can directly describe the subject, environment, action, and camera changes to generate a video. Prompts work best when written as a clear shot, such as where the subject moves, how the camera approaches, and what atmosphere the lighting conveys; focus on a single action to make it easier to compare different creative options.
When you already have product, character, or scene images, you can enter the image-to-video workflow through image_url, then use text to describe the action you want to happen. The image provides the visual starting point, while the prompt adds motion intent, making it suitable for exploring short clips such as a character turning, environmental changes, or camera push-ins, rather than redefining the entire scene from text.
Video generation can use asynchronous submission: first obtain a task_id, then check the status or receive a completion callback. The result provides video_url, allowing applications to handle submission, waiting, previewing, and downloading separately. For content production systems, this task-based delivery is easier to organize than keeping users on a waiting page throughout the process.
Start with specific tasks to find where the model can be most effective.
Use a static product image as input, add background atmosphere, motion direction, and camera distance, and generate candidate short videos for presentation. Start by trying one clear action, then pass suitable clips to the editing stage to add selling-point text, branding, and music. This is well suited for teams that need to expand existing visual assets into dynamic content.
Rewrite a shot from a script as a description of the subject, scene, action, and camera movement, then create a watchable draft through text-to-video generation. The deliverable can be used to discuss pacing, composition, and action intent, without requiring complete live-action footage in the early stage; different descriptions can be generated separately to help creators select a more suitable direction of expression.
Use existing illustrations or character images as a starting point, specify subtle movements, viewpoint changes, or environmental motion, and create dynamic proposals. This is suitable for exploring formats for event visuals, character showcases, and social content. Details that need to be retained in the final clip can be specified clearly in the prompt and checked item by item when selecting clips for subject form and motion results.
Choose based on task complexity, input materials, and expected results.
If you need to retain both text-to-video and image-to-video modes, and each clip is within the 1–15 second range, you can choose this model. grok-imagine-video-1.5-fast:reverse supports durations of 6–30 seconds, making it better suited for tasks requiring longer clips. The two do not use the same invocation ID; when switching, adjust the duration settings accordingly rather than simply replacing the name while retaining all parameters.
grok-imagine-video:official also provides text-to-video and image-to-video modes and is another public invocation entry point; grok-imagine-video-1.5:official supports image-to-video only and requires an image. When you only have a text concept, you can start directly with this model; when you already have an image and are ready to compare other versions, you can test each separately using the same material and action description.
From a small-scale task to production integration.
Clarify the goal, required inputs, and output requirements, using real business examples as a starting point.
Open the trial page, confirm the parameters supported by this entry point, then submit a small-scale task to review the results.
Keep the complete model ID, use the request format specified in the documentation, and confirm billing rules on the Pricing page.
Before formal use, understand the output quality and capability scope.
Answers to common questions about using grok-imagine-video:reverse.
Yes. This model supports text-to-video; simply provide a prompt to describe what you want to generate. It is recommended to specify the subject, scene, action, and camera movement, and organize the information around a single shot first. When calling it, explicitly enter grok-imagine-video:reverse to avoid using another default model.
When providing image_url, prompt can be left blank. However, if you want the subject to move in a specific direction, or want the camera to push in or rotate, adding an action description makes it easier to express your intent. The image provides the visual starting point, while the text explains how you want the scene to change; they serve different purposes.
This model supports durations of 1–15 seconds, with a default of 6 seconds, and is not suitable for directly submitting a 30-second task. For longer single-shot videos, consider grok-imagine-video-1.5-fast:reverse, which supports durations of 6–30 seconds; for multiple shots, you can also generate them separately and edit them together.
task_id is a task identifier, not a video URL. When using async, you can check progress through POST /grok/tasks; when callback_url is set, you can wait for the completion notification. After confirming that the result status is succeeded, read video_url; pending means it is still being processed.
:reverse is a suffix for the public call ID used to distinguish entry points, and should not be treated as an independent vendor model name. When using this entry point, retain the full ID. It shares the basic text-to-video and image-to-video creation methods with the :official entry point, but this does not mean that all available parameters are exactly the same.
Model information · Updated: 2026-10-01. For calling parameters and billing rules, see the API and pricing sections.
Start with a clear goal and judge from actual results whether it suits your work.
