Grok Imagine video and motion, without burning the budget

Monday, 9:06 a.m. The launch page still needs a loop. Someone generated twenty Imagine clips overnight “to see the vibe.” You scrub them at 1.5× on a laptop with 11% battery. Clip 3 is a violent zoom into the mug. Clip 9 morphs DOCK 4 into D0CK mid-pan. Clip 14 has a faint Sony-shaped mark on a speaker that was never in the still. Clip 20 is a six-second earthquake. None of them can go on the page. The 10:30 calendar invite still shows hero_still_v3.jpg because a still, at least, does not spasm.

Motion is a second product. Do not sprinkle it on a PNG to look busy. This is Part 23 of the Grok series. Part 21 was a brief from zero. Part 22 was edits that stick. Here you keep the still if it would already ship, then ask for one short clip with one camera move. Analytics Made Simple is educational. We are not an xAI partner. Duration, audio, and prices move. Re-check docs the week you click generate.

When a still is the asset and motion

When a still is the asset and motion is optional garnish
Text-to-video vs image-to-video, and why image-to-video is the work default
One camera move, no extra plot, no extra props
Length (seconds to about 15 seconds), audio, and cost hedges
A toy 6-second spec you can paste after you lock the still

Still vs motion

A still has one job: survive a pause. People screenshot. Calendar apps freeze on frame one. Slack unfurls a JPEG. If the still is ugly, the clip will not save it. If the still is good, you may not need a clip at all. Ask the rude question before you spend video credits: would you ship this JPEG if video did not exist? If no, go back to Part 21 and Part 22. Do not animate a misspelled sleeve and hope motion hides the letters. Motion makes letters worse.

Use motion when the page or story needs time: a slow push toward a product, steam that already existed in the still, light moving across a table. Skip motion for a quarterly deck cover, a print hallway poster, or a thumbnail that will be paused 90% of the time. A looping earthquake behind a KPI does not make the KPI truer.

Imagine motion path: lock a shippable still, pick image-to-video, one camera move, short duration, inspect, then keep or kill. Do not sta…

Twenty clips is a binge. Each clip has a cost (API video is billed with duration and resolution in the mix; consumer quotas change). Twenty vibe tests also train your eye to like noise. Pick one still. Pick one move. Generate one or two clips. Inspect. Keep or kill. If both fail, change one line in the spec, not the entire story.

Text-to-video vs image-to-video

Imagine’s video family (1.5 era as of writing) can start from words or from a still. Text-to-video invents the first frame. That is fun for a private mood board. It is a poor default for work because you already fought for a mug, a teal, and a sleeve. Image-to-video takes the still as the starting picture and asks the model to move. That is the path that respects Part 21 and Part 22.

Docs also describe other video jobs (edit an existing clip, reference-to-video, extension). Hedge: those routes exist for builders. For a launch page on Monday, you need image-to-video and a duration you can afford. The API example model id in current Imagine docs is in the grok-imagine-video-1.5 family. Names move. Confirm the live picker.

Image-to-video still drifts. It can invent a hand. It can boil the type. It can add a logo on a device. Your inspect pass is the same as for stills, plus motion-specific fails: warping type, extra objects fading in, a camera move that fights the product. If the still had empty chairs, the clip should not grow a photoreal coworker in second 3.

Keep the shot simple

One camera move. Write it like a camera operator on a tiny set, not like a movie treatment.

Slow push-in toward the mug (10% closer, no orbit)
Locked tripod, steam rising, nothing else
Gentle pan left across the table, stop before the plant exits

Refuse the rest. No crash zoom. No orbit around the product. No “then the camera flies out the window into the city.” No extra characters. No new props. No title animation. If the still has on-image type (you already know that is the hard part), tell the model to hold the letters still. If you can, use a still with no type and put words in the page HTML or the slide. Type that must live inside video is type you will inspect frame by frame.

Simple Imagine shot card: locked still, one move (slow push-in), hold type, no new props, duration about 6 seconds, inspect for warp and …

Audio is a maybe. Some Imagine video paths can add native sound. Confirm the live toggle. For a muted autoplay on a landing page, you often want silence anyway. Browsers mute autoplay. A sudden whoosh on a looping hero is how you annoy people with headphones. If you need sound, write it as “soft room tone, no voiceover, no licensed song.” Do not ask for a celebrity voice. Do not ask for a famous chorus.

Rule of thumb: If you need a storyboard with more than one arrow, you are making a film. Imagine is making a clip.

Length, audio, and cost hedges

Clips are on the order of seconds to about 15 seconds. That is the honest ceiling to plan around, not a promise of a 15.00 second slider every week. API samples in the Imagine docs use durations like 12 seconds. Confirm live limits and the max in your picker. Do not write a brief for a 90-second brand film. You will get a smear, a bill, or both.

Cost: image generation is typically a flat per-image fee on the API. Video is per-second, and resolution joins the bill. Consumer apps use plan quotas that change. Hedge every number. The work habit is the same on any plan: decide duration before you generate. Six seconds is enough for a hero loop. Eight is plenty. Twenty clips at 12 seconds is how a vibe test becomes a line item someone has to explain.

Resolution: a landing-page loop does not need a cinema master. Pick the size the page will play. Then watch it on your phone in bright light. Compression will eat fine type. Another reason to keep letters out of the clip.

If you extend a clip (some API routes can continue from the last frame), you are stacking drift. Extension is a builder feature. For Monday’s page, cut. Do not stitch five extensions into a “mini episode.”

Worked 6-second spec

Toy: the Dock 4 mug still from the last two posts, 16:9, no people, sleeve reads DOCK 4, glaze in the teal family. You want a quiet loop for a launch block that already has an H1 in HTML. Calendar fallback stays the JPEG.

Do this in order. Do not skip the still lock.

Open the still at 100%. Confirm spelling, no extra fingers, no fake marks.
Decide duration: 6 seconds. Write it down so you do not bump it “a little longer” six times.
Choose image-to-video. Attach the still.
Paste one move. Generate once, maybe twice. Watch at 1×, not 1.5×.
If type warps or a logo appears, kill the clip. Keep the still. Do not generate clip 3 through 20.

Image-to-video from this still. Duration: 6 seconds. 16:9.
Camera: slow push-in, about 10 percent closer, locked horizon, no orbit, no tilt, no handheld shake.
Motion: faint steam from the mug if it already looks warm. Plant leaves still. No new objects. No people. No extra hands.
Type: hold the sleeve letters exactly as DOCK 4. Do not morph, flicker, or add words.
Marks: no logos on the laptop, mug, or speaker. No brand names.
Audio: silence.
Refuse: crash zoom, earthquake, city fly-through, photoreal staff, celebrity voice, fake Sony or apple-ish marks.

What to do with the output (toy checks, not a screenshot):

Starting still	Motion request	Fail if
Dock 4 mug, 16:9, sleeve DOCK 4, no people	6s slow push-in, steam only, silence	Type morphs, extra mug, fake mark, crash zoom
Same still, type already wrong (`DOK 4`)	Any motion	You started video before fixing the still
Empty conference room, no product	Pan across the table	A photoreal coworker fades in
Product cutout on cream	Locked tripod, gentle light shift	A marble store appears behind the cutout
Poster with three lines of type	You insist on video anyway	Letters boil; use the PNG on the page instead

Ship rule for the toy launch: if the clip fails twice, the page goes live with hero_still_v3.jpg and a caption. You still have a Monday. You do not have twenty unusable MP4s and a dead battery.

Ship the still first

Mistake	Fix
Twenty vibe clips before a shippable still	Lock the JPEG. Then one clip.
Text-to-video for a product you already designed	Image-to-video from that still
Three camera moves plus a plot	One push, pan, or locked steam
Animating misspelled type	Fix letters in stills (Part 22), then consider video
Assuming 90-second films	Plan seconds to about 15s. Confirm live max.
Autoplay with a whoosh and a voice	Silence for muted heroes; no celebrity audio
Leaving the invite on a JPEG after a good clip	Fine. A still fallback is a feature.

Ship the still first (light)

Take one still you would already publish. Write a six-second spec with one move. Generate at most two clips. Watch at 1× on your phone. Keep one or keep the JPEG. Next: Grok Imagine rights, deepfakes, and common fails. That post closes this Grok series. Hub: analyticsmadesimple.com/series/grok. Video docs to re-check: Imagine overview.

Ship the still first 3

Ship the still first. Motion is optional.
Default to image-to-video from a locked frame, not twenty text-to-video tests.
One camera move. Hold type. No new people or marks.
Plan seconds to about 15 seconds. Confirm duration, audio, and price the week you run it.
Two tries, then the JPEG. A calm still beats a looping earthquake.

Sources

Research and further reading used for this article:

xAI docs: Imagine overview (text-to-video, image-to-video, duration notes up to about 15s, video family, pricing shape)
xAI: Imagine Image 2.0 (stills, Quality Mode, world-building stills that can feed video later)
Grok Imagine (consumer studio: images and short video)
xAI API (Imagine models billed apart from chat subscriptions)
xAI Console (keys if you script video; do not paste real keys)
Analytics Made Simple: Grok series
Analytics Made Simple: Learn

Hands-on data strategist, analytics engineering lead, and educator. Writing practical, no-fluff guides to help everyday teams, analysts, and engineers master SQL, AI systems, and modern data architectures.

About Jose → More articles → Get in touch

Keep going

Same lessons in your feed

Short diagrams, hooks, and weekly tutorials on Substack, Instagram, X, and Facebook.

@analyticsmadesimple Follow @analyticsmadesimple Follow @AnalyticsSimply Follow Analytics Made Simple Follow