Grok Imagine

Name: Grok Imagine
Author: xAI

byxAI

Live-action clips that arrive with their own dialogue, room tone and effects, from a prompt, a first frame, or up to seven references

Roll the shot

How Grok Imagine works

One shot per generation: write the scene and its sound, decide whether images lead the frame, then draft small and render the keeper larger.

Grok Imagine workflow connecting a detailed transit-pod prompt to the video generator.

Frame the shot and the sound

Write one scene with a camera move, then add an Audio: line for dialogue in quotes, ambience, or 'no music'.

Grok Imagine set to version 1.5, six seconds, 16:9, and 720p with the generated transit-pod video.

Cite your reference stills

Add one image to open on that frame, or two to seven and cite them as @Image1 through @Image7, each with one clear job.

Grok Imagine result showing a transit pod crossing reflective salt flats at dawn.

Draft short, render the keeper

Pick 1.0 or 1.5, set 1 to 15 seconds, preview at 480p, then re-render the keeper at higher resolution.

What Grok Imagine is good at

Built-in sound, reference-driven scenes and calm, believable camera behavior make this a fast tool for short, finished-feeling clips.

Picture and sound in one pass

Dialogue with lip-sync, room tone, effects and music are generated together with the footage, so a character can speak a quoted line and the clip arrives finished. Name the ambience or say 'no music' to steer it.

Seven-image reference blending

Feed two to seven images and cite them as @Image1 to @Image7 to blend character, location and prop into one coherent clip. Reference mode runs 1 to 10 seconds, and v1.5 gives the best identity and setting consistency.

Draft on 1.0, finish on 1.5

Version 1.0 is the cheaper previous generation for quick motion tests; v1.5 adds 1080p for text and image modes and stronger weight and momentum. Draft at 480p, then spend the budget on the keeper.

Steady, believable camera moves

Restrained, realistic camera behavior responds to concrete filmmaking terms like dolly in, tracking shot or drone fly-through. Keep to one move per clip and 1 to 15 seconds of runtime; 5 to 8 seconds holds detail best.

Made with Grok Imagine

A spread of single-beat shots (ads, a duel, a gallery room, a macro product moment, a watch) each carrying its own soundtrack.

Brand commercial with athletic motion and punchy foley

Cinematic duel, one clean beat of action

Interior space with shifting natural light

Tactile product moment with close-miked sound

Luxury timepiece with mechanical acoustics

What people build with Grok Imagine

Social creators, ad teams, directors, product brands and visualizers use it when they need a quick, sound-included clip rather than a long, choreographed sequence.

Talking-character social creators

Make vertical 9:16 clips where a character delivers a quoted line with lip-sync and ambient sound already baked in. Draft cheap at 480p, re-roll fast, then render the take you like.

Ad and teaser makers

Produce square and vertical ad shots, single-action teasers with sound included. v1.5 gives you 1080p for the final cut after short drafts.

Directors building animatics

Assemble scenes from character, location and prop references, then chain one-action shots into a fast animatic. Each clip carries sound, so pacing can be judged before a crew is booked.

Product teams animating stills

Bring a hero product photo to life as the opening frame, with a slow push-in, a light sweep and a quiet audio bed. Describe only the motion; the image supplies the look.

Architectural visualizers

Turn an interior or exterior concept into a steady fly-through or light-shift clip with ambience, to test mood before a full render pipeline.

Grok Imagine specs

Plans and credits

Versions: Grok Imagine 1.5, Grok Imagine 1.0
Inputs: Prompt (text), Images (up to 7 images)
Output: video
Duration: 1–15 seconds
Resolution: 480p, 720p, 1080p
Aspect ratios: 16:9, 4:3, 3:2, 1:1, 2:3, 3:4, 9:16
Cost: 69–5,166 credits per run

Grok Imagine questions

Version 1.5 is the newer default and adds 1080p for text-to-video and image-to-video, with better motion physics (weight, momentum). Version 1.0 is the cheaper previous generation, best for drafts, and has no 1080p. Costs on 1.5 climb steeply at long durations and high resolution, so draft at 480p first.

Use text-to-video for quick ideas from a prompt alone, image-to-video to animate one still as the opening frame, and reference-to-video to blend two to seven images (character, location, prop) cited as @Image1 to @Image7. Reference mode is capped at 10 seconds; text and image modes run up to 15.

Yes, it generates picture and synchronized sound in one pass, including lip-synced dialogue, ambience, effects and music. Put spoken lines in quotes and add an Audio: line, or say 'no music', to steer the track.

Text and image clips run 1 to 15 seconds; reference clips run 1 to 10 seconds. Clips of about 5 to 8 seconds hold identity and detail best, since fine detail can drift toward the 15-second end.

It struggles with long multi-step action, dense crowds, complex hand work, and precise on-screen text or logos, where ghosting and warped anatomy can appear. Keep to one shot and one action per generation, and chain clips for sequences.

It was created by xAI. This node exposes text-to-video, image-to-video and reference-to-video modes, at 480p, 720p and 1080p (1080p on v1.5 text and image modes).