SpaceXAI: Grok Imagine Video

Grok Imagine Video is the classic edition of the Grok Imagine video family. Its current manufacturer documentation covers generating a new clip from a prompt, animating a starting image, using visual references and changing or continuing an existing video. These workflows produce different artifacts: a new scene, an edited version of supplied footage or one video containing the original footage followed by new content. Choose the workflow from the material already available and the part of the shot that needs to change.

For a new text-to-video shot, describe the subject, environment, visible action, camera and sound. Standard generation supports 1–15-second clips, with 480p or 720p output and generated audio. Square, landscape, portrait and wider ratios are documented, including 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9 and 5:2. An original brief might ask for a paper lantern moving gently above a quiet courtyard as the camera tracks sideways. The useful review checks whether the motion, framing and ambient sound support that one shot.

In image-to-video, the supplied still is the first frame. Prepare it with the intended crop, subject and visible details, then describe the action that follows. The output follows the image’s aspect ratio. A starting still gives an exact opening composition, while reference-to-video uses images to guide the people, products or environment in a newly composed scene. Keeping those two roles distinct helps avoid an unintended opening when the goal is to place a known object into another setting.

The classic model accepts up to seven reference images per reference-guided request. That mode is limited to ten seconds and 720p. Give each reference a clear role and identify it by its position in the list, beginning with <IMAGE_0>. For a product vignette, one reference can identify the object and another the desired location. Keep the brief focused on the object’s placement and the shot’s main action. The classic reference mode uses its own configuration; the full 1.5 edition supplies the additional selected-voice and pinned-frame controls.

Video editing takes an existing clip and a prompt describing the change. The maker accepts MP4 sources with supported codecs such as H.264, H.265 or AV1, up to 8.7 seconds long. The edited result keeps the source duration and aspect ratio, and follows its resolution up to a 720p cap. Custom duration, aspect-ratio and resolution controls do not apply to this edit workflow. Prepare a source that already has the composition and timing needed for the final use, so the prompt can concentrate on the intended visual modification.

An original editing brief might change a scarf’s colour while retaining the person’s face, movement, background and camera. State those preserved elements explicitly alongside the change. The manufacturer describes retaining the rest of the source as closely as possible, which is a useful objective for the review rather than a guarantee that every pixel remains identical. Compare the result through the whole clip, especially where the object moves, crosses another object or becomes partly hidden. The final artifact is the edited clip, not merely a still showing the new colour.

Video extension continues from the last frame of an existing MP4. The documented source length is 2–15 seconds; the added portion is 2–10 seconds, with a six-second default. The extension duration concerns only the new portion. A ten-second source followed by five seconds of continuation produces a fifteen-second video containing both parts. The result preserves the source aspect ratio and resolution within a 720p cap. Prepare the source’s final moment and describe what should happen next, such as the subject turning toward an approaching object while the camera continues its current move.

Review an extension around the boundary between the source and the new content. Check subject identity, movement direction, camera pace, light and sound so the continuation fits the preceding action. If a clip ends at a visually awkward point, make that boundary an explicit part of the next-shot plan. Longer scene assembly is an editing decision: the model’s generation, reference, editing and extension duration limits are separate, and an extension’s total result can be longer than a standard new-generation clip.

The maker’s video workflow is asynchronous. A request identifier tracks a job until a completed result supplies the video location and duration; output locations are temporary. Retain the playable result and its source materials when preparing a production edit. Full Grok Imagine Video 1.5 offers native 1080p generation and richer visual, voice and frame anchoring, while Lite serves simple prompt or starting-image work. The classic edition remains useful when the starting material is already a video that needs a directed alteration or continuation.