Can 1.5:official generate videos using text only?
No. This model uses image-to-video and requires image_url. Text prompts are used to supplement the actions, camera work, or scene changes in the image. If you do not yet have image assets, use grok-imagine-video:official or fast:reverse, which support text-to-video.
Do I still need to write a prompt in addition to the image?
Image-to-video can be used without a prompt, but adding one is recommended when you have clear motion goals. Descriptions should focus on how the subject moves, how the camera changes, and what happens in the scene. Avoid requesting too many conflicting actions at once, so the goal of the short video is easier to assess.
How long and at what resolution can videos be generated?
This model supports 1–15 seconds, with 6 seconds by default; available resolutions are 480p, 720p, or 1080p, with 480p by default. It is suitable to first create short drafts to check movement, then generate higher-resolution candidates for satisfactory ideas. The specific footage still needs to be reviewed in the end.
How do I retrieve asynchronously generated videos?
After setting async to true, save the returned task_id, then query the results through /grok/tasks; you can also provide callback_url to receive completion notifications. Only after the status reaches succeeded and video_url has been obtained should the result be considered a downloadable video asset.
How do I ensure I am calling this model?
When submitting a request to /grok/videos, explicitly set model to grok-imagine-video-1.5:official and provide image_url at the same time. Do not omit the model or use a name containing preview; retain task_id and trace_id to help link assets, task status, and generation results.