Grok Video Prompts: A Six-Part Guide

1. Subject: who or what, with two or three visible details (age, clothing, material). 2. Action: one main movement, with where the hands or the object go. 3. Camera: the lens or framing and one move, such as a slow push-in. 4. Light: where it comes from and its colour. 5. Sound: what is heard, including any spoken line in quotes. 6. Ending: the last frame, so the clip lands instead of drifting.

xAI's docs say text-to-video on Grok Imagine 1.5 first draws a frame from your prompt and then animates it, so the first sentence, the subject and the setting, carries the most weight. We found no official xAI prompting guide; the structure above is how Reviral writes the prompts behind its own example clips.

Product: A matte black water bottle stands on a wet granite counter; a hand lifts it and condensation runs down the side. Macro lens, slow push-in. Soft window light from the left. Sound: a cap twisting open. Ends on the logo facing the lens.

Person: A woman in her twenties in a tangerine trench walks toward the camera across a wet crosswalk at dusk. 50mm lens tracking backward. Amber streetlight from the left. Sound: boot heels on asphalt. Ends as she stops one step from the lens.

Talking: A barista leans on the counter and says, "Try it iced." Medium close-up, still camera. Warm cafe light. Ends on her smile.

Reviral has no Grok model; the same six-part prompt runs on the models in Cinema Studio, such as Seedance 2.5 (4 to 30 seconds, up to 30 reference pictures) or Wan 3.0 (2 to 30 seconds). The free AI Video Prompt Generator writes a shot-by-shot prompt from a short brief.