A short clip is a tool. A moving likeness of someone who did not agree is a problem. This lesson is the clip you are allowed to make.
Why this matters
Video spends more of your weekly Grok usage than a paragraph. It also spends more legal and reputational credit. A six-second push-in on a diagram can explain a hold point. A six-second clip of a coworker’s face, a student’s face, or a client’s yard “brought to life” can become something you cannot take back.
xAI’s Acceptable Use Policy (current as of August 2026) forbids, among other things: undressing or nudifying real people; deceptive impersonation; pornographic likenesses; depicting someone in a false light; and any sexual content involving minors. Enabling any “NSFW” control does not turn those rules off. This course adds a work rule on top: if you do not have consent to use the person’s image, you do not generate them, still or moving.
You will make one short, work-useful clip from a still you own or from the invented diagram in the last lesson. You will write the consent check first. If the check fails, you change the still. You do not generate “anyway.”
Teach
On grok.com, Imagine can start from text or from a still. For work, start from a still you already approved. Text-to-video invents extra people and extra logos. Image-to-video has a first frame you can inspect.
A useful work clip is six to ten seconds. That is enough for:
- a slow push toward the HOLD label on a diagram
- wind in grass on a site photo you took, with no faces
- a product or tool on a bench, rotating slightly, if you own the object and the photo
- a process: one arrow drawing itself across a four-panel graphic
It is not enough for a brand film, a talking spokesperson, or a deepfake greeting from the mayor.
Write the video prompt like a camera card, not a screenplay:
- First frame. Which still, and why you have the right to use it.
- What moves. Camera only, or one element (arrow, water, clouds).
- What stays still. Labels must remain readable. Faces — if any — should not exist.
- Duration. Six seconds unless you have a reason.
- Audio. If the product offers sound, say “no speech, no one else’s voice.” Do not clone a voice. Do not put words in a real person’s mouth.
- Must not appear. Same ban list as the still, plus “no new people walking into frame.”
Likeness and consent — read this before you hit generate:
- Who is in the still? If the answer is anyone other than you, an invented/faceless figure, or nobody, stop.
- Did they agree to this use? Written is better than a shrug in the hallway. “They posted the photo on Facebook” is not consent to animate them.
- Could a reasonable person think this is that person speaking or acting? If yes, you are in impersonation territory. Stop.
- Is anyone 17 or under in the frame, including in the background? Do not generate. Do not crop and hope.
- Is the request sexual, nude, undressed, in underwear, or “in a bikini for fun”? Do not generate. Not of a stranger, not of a coworker, not of a public figure, not of yourself for this course.
- Would you put this clip on the company letterhead? If no, it is not a Grok Mastery artifact.
If you are depicting yourself, say so. That is allowed here. Keep it work-useful. You are still responsible for where you post it.
Watermarks stay. Provenance stays. If you publish the clip, say it is AI-generated. Do not present it as field footage of a real event that did not happen.
Plans differ. SuperGrok and usage limits may cap resolution or volume. A 480p six-second clip that does the job beats a 720p clip you cannot defend. If video is blocked on your plan, attempt the generate so the block is real, keep that message, and finish the camera card and consent check in chat. You still opened Imagine.
Buttons will move. The consent questions will not.
Worked example
The field lead still has the Thursday inspection diagram — no people, four labels. She uploads that still into Imagine video and writes:
First frame is this diagram. Six seconds. Slow push-in toward the HOLD box at the transformer pad. Labels stay sharp. No people enter. No logos. No speech. Overcast, quiet, documentary pace — not cinematic explosion.
First clip: Imagine invents two workers walking in from the left. She does not “fix their vests.” She reruns: “Same camera. No people at any frame. Cropped if needed.” Second clip is clean. She drops it behind the spoken tailboard. She never used the crew photo from the parking lot.
She also fills a consent card before the first generate. Line 1: “Who is in the still? Nobody.” That is the whole check. If line 1 had been “Dana from the warehouse,” the card would have stopped her.
Your turn
Open grok.com Imagine. Use the still from the last lesson or a photo you took of a place, object, or diagram that contains no other person. New chat thread for the cards. Do the consent check before you generate.
Exercise a6-l02-e1
| Field | Value |
|---|
| whereToDo | grok.com Imagine (video) |
| time | 20 min |
Copy-ready prompt
I am in Grok Mastery (Russell Tech), Pathway A, lesson A6 — short video and likeness.
Do not write a video prompt yet. Interview me, one question at a time:
1) What still will I animate (last lesson's diagram, or a photo I took)?
2) Who, if anyone, is visible — face, body, reflection, background? If the answer is anyone but me or nobody, tell me to pick a different still and stop.
3) Do I have consent to use this still for an AI video? If it is my photo of a place/object with no people, say "property of learner." If it is my face, I must type "I consent to use my own likeness for this course clip."
4) What work job will the six-second clip do?
5) What should move, and what must stay readable?
Then write two cards.
CONSENT CARD
- Who is in the still:
- Rights/consent line (must be "no person" or "my likeness, I consent"):
- Minors in frame: yes/no (must be no):
- Sexualized or undressing request: yes/no (must be no):
- Impersonation risk: yes/no (must be no):
- Letterhead test (would I put this on company letterhead?): yes/no:
- Go / No-go:
CAMERA CARD (only if Go)
- First frame:
- Duration:
- What moves:
- What stays still:
- Audio: no speech, no cloned voice
- Must not appear:
- Imagine video prompt (one pasteable block):
- If people appear in take 1, take 2 prompt:
Session stamp: today's date and "These cards were written by Grok in this thread."
If Consent is No-go, do not write a video prompt. Tell me to change the still.
If the card is Go, generate the clip. If take 1 adds people, run the take 2 prompt. Then send:
Take 1 result: [no people / people appeared / blocked by plan or policy].
If you wrote a take 2 prompt, I ran it: [result].
Refresh the session stamp only. Do not loosen the consent card.
Artifact to paste
Paste, in this order:
- The Consent Card and Camera Card (or the No-go and the replacement still’s cards) with session stamp.
- Result line: clip generated (what moves, duration), or the exact Imagine block / limit message.
- One sentence you wrote: the work use of the clip, and the words “no third-party likeness.”
Do not upload a clip of a real person who is not you. Do not upload anything sexual. A text description of the clip is enough for the course desk.
Rubric
- Pass: Consent Card completed before generate (or a documented No-go plus a second, clean still). Rights line is only “no person” or “my likeness, I consent.” No minors, no sexualized request, no impersonation. Camera Card names duration and what stays readable. You attempted Imagine video, or captured a real plan/policy block. Stamp present. Final sentence includes “no third-party likeness.”
- Fail: You animated someone else’s photo. You generated first and filled the consent card after. A talking-head deepfake, even “as a joke.” Any undressing or sexualized prompt. You skipped Imagine entirely. You cloned a voice. No session stamp.
Common failure
The common failure is “it’s only six seconds.” You animate a group photo because the motion looks impressive. Someone in the back row never agreed. You post it. They find it.
The fix is the first line of the consent card. If anyone but you is in the still, pick another still. Do not crop a child out and call it clean. Do not test the NSFW toggle “to see what happens.” This course will fail that artifact on sight.
The second failure is a movie treatment: weather, extras, a crane shot, a speech. Usage gone, labels unreadable, new faces invented. One camera move. One job.
What’s next
You can make a still and a short clip without stepping on someone. Next module: everyday operating system — email and a week plan, then learning a real topic without handing Grok your thinking.
