SpaceXAI: Grok Imagine Video 1.5

Grok Imagine Video 1.5 is the full generation edition of the Grok Imagine video family. The manufacturer emphasizes motion, physical continuity and sound effects, ambience and speech generated with the picture. Text-to-video and image-to-video render at native 1080p, while reference-guided generation supplies more control over who or what appears in a scene. The useful result is a short video with its audio, such as a product presentation, a character exchange or a precisely staged transition between visual anchors.

For text-to-video, describe the shot’s subject, setting, action, camera movement, lighting and sound. The manufacturer explains that 1.5 generates a starting image internally and then animates it; the intermediate still is not returned. For image-to-video, a supplied image becomes the first frame and the prompt describes what happens next. Prepare that image with the desired framing and visible details already present. The resulting image-to-video clip follows the source image’s aspect ratio, so a vertical delivery needs a suitable starting crop.

Reference-to-video gives images a different role: they guide people, objects, clothing or locations without automatically fixing the opening frame. The documented limit is up to 14 reference images, with output up to 15 seconds at 720p. Assign each image a clear role in the scene and identify it by its position in the reference list. In the manufacturer’s prompt syntax the first reference is <IMAGE_0>, followed by <IMAGE_1> and so on. When an opening image is also pinned, it takes the first index and the separate references start with the next one. This ordering matters when a brief distinguishes a person, a product and a setting.

An original reference brief might place a character beside the chair supplied in one image, inside the room supplied in another, with a slow move toward the chair’s material detail. Keep the subject and key product easy to distinguish in the references. The maker advises giving each reference one role and a fixed place, keeping the cast small and props static when using many inputs. Review whether the intended object’s shape and markings carry into the scene, as well as whether the camera reveals the detail requested in the brief.

Reference-to-video can also assign up to three preset voices. Identify the intended speaker and its voice separately in the prompt, using the documented positional voice references, and provide the exact dialogue that the speaker should say. The model supports lip-synced speech. For a short exchange, write lines that leave room for the reply and inspect both speaker attribution and mouth movement in the resulting clip. A voice reference selects the speaking identity; the camera and scene directions still need to describe how the characters are shown.

The first- and last-frame controls anchor the endpoints of a shot, and up to four keyframes can pin images at intermediate times. The maker’s keyframe guide places those times strictly inside the clip on a one-third-second grid; two anchors that resolve to the same time slot are rejected. Prepare anchors that belong to the same scene and describe the motion between them. For a product reveal, the opening might show the closed display and the final frame the revealed object. For a loop, the same image can serve as the first and last frame. Listen and watch through the transition rather than judging only the pinned images.

Ordinary text- and image-to-video support 1–15-second clips and 480p, 720p or native 1080p output. Reference-to-video is limited to 720p. Text-based framing includes 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9 and 5:2. Choose the duration and shape for the action and destination. The maker generates audio by default and documents a silent-video option. A brief can specify the spoken line, nearby effects and distant ambience so sound contributes to the same scene instead of becoming an unspecified background.

The documented generation flow is asynchronous: it returns a request identifier, then a completed result with the clip location and duration. The example result points to an MP4 video, and its hosted output location is temporary. Preserve the completed clip and the input brief when preparing an edit. Check continuity, object persistence, motion, framing and the sound against the intended shot. An intermediate processing record is useful for tracking the work, while the playable video is the artifact to review.

Version differences affect the plan. Lite handles straightforward text or single-image animation, with its 1080p option upscaled from 720p. Full 1.5 adds reference images, preset voices and pinned frames. The classic Grok Imagine Video edition provides editing and extension of existing videos. The release’s separate Video 1.5 Fast speed figures concern that named variant, so they do not define a fixed completion time for this model. For a longer sequence, stage coherent short shots and review character, product and sound continuity when arranging the clips together.