Grok Imagine will refuse prompts that other tools accept. That is not a puzzle to hack. It is a filter. The useful move is to say the same picture in calmer words. This page stays safe for work and does not teach anyone to sneak past a rule.

Quick answer
Remove names of real people and brands. Remove violent or suggestive words. Use a short, ordinary scene. Start the line with safe for work if you like. Try once. If it fails, pick a different scene.

Why blocks happen
Filters look for famous people, sexual content, graphic harm and some brand names. They also trip on wording that sounds like a fight or a trick. You cannot see the exact rule. You can see that a calmer sentence often passes.

A rewrite pattern
Before: make me look like a famous actor in a bloody action poster. After: a cinematic portrait of me on a dark street, wet asphalt, no text, safe for work. The look stays. The triggers go.
Use the generator
The Grok prompt generator always writes safe-for-work stills or clips with no famous people and no logos. The Grok hub is the same rule on library cards.
What not to do
Do not misspell banned words. Do not ask the model to ignore its rules. Do not upload someone else's face. Do not keep retrying the same blocked line twenty times.
If you need a different tool
Gemini and ChatGPT have their own rules, but a plain portrait often works there. For video, try Veo or Sora with the same calm shot list. For local control, use Stable Diffusion and accept that you own the whole stack.
Field notes
Filters change without notice. A line that passed last month can fail today.
Clips are often stricter than stills. Keep clip actions small.
If you need a couple photo, use two of your own pictures and calm wording. See the couple generator.
Horror in this library is atmosphere, not gore. The horror collection is written that way.
We will not list secret trigger words. That list would be wrong next week anyway.
A block is not always your fault. Accounts and regions differ.
Keep a private log of rewrites that passed, without the blocked originals if they were sensitive.
If you are writing for a brand, drop the logo request and describe a blank mark.
Grok Imagine Says Content Moderated is a method you can repeat. Write down the date, the tool name you saw, and the one line that worked. That note will outlast a product rename.
Example images on PhotoPrompts are illustrations of a brief. They are not proof that a named model made the picture. The badge on a card reads Works with, never Made with.
If a menu in your app does not match a sentence here, follow the menu. Tools move buttons. The brief you write is the part that stays useful.
Do not invent a statistic to make a prompt sound scientific. If we have not measured a difference, we say so and we do not invent a number.
Keep other people's faces out of your uploads unless they have agreed. A good prompt is not a licence to use someone else's likeness.
When a result is close, change one word or one setting. When a result is wrong in every way, start a new chat and shorten the brief.
Save the text, not only the picture. Pictures get lost in chats. A sentence in a note can be pasted again next month.
If you want a person to look at a stuck result, use the contact page and send the prompt, the model and a screenshot. Mention that you followed Grok Imagine Says Content Moderated.
A weekly habit beats a binge. Ten careful attempts with notes will teach you more than a hundred un-saved retries on a Sunday night.
If two tools disagree, keep both results for a day before you delete one. A second look in daylight is a better judge than a late-night refresh.
Print or screenshot your best brief and keep it near your desk if you do this for work. Templates save you from rewriting the same lock line badly.
If you only remember three things
- Rewrite in ordinary, safe words.
- Drop names and brands.
- Do not try to trick the filter.
Worked example
Here is the same idea written two ways. The second one is the one to paste.
Before: Make me look like a famous actor in a movie poster with a fight.
After: A cinematic portrait of me on a dark street, wet asphalt, camel coat, no text, safe for work.
The second line describes a picture. The first line asks for a person you do not have the right to use.
A practice you can finish today
Set a timer for twenty minutes. Do not chase a perfect picture. Chase a note you can reuse.
- Take a blocked idea you actually had, if any, and rewrite it in one calm sentence.
- Run it through the Grok generator options instead.
- If you do not use Grok, practise the rewrite anyway. Other tools have filters too.
- Save the calm sentence.
- Delete the version with the name.
Write down the line that worked, the model name you saw, and the date. That note is worth more than a long prompt library you never open.
Mistakes that waste attempts
Most wasted attempts come from changing too much at once, or from asking the model to guess.
- Misspelling a name to sneak it through.
- Twenty rapid retries.
- Uploading a face you do not have permission to use.
- Adding more adjectives after a block.
- Writing a how-to for bypassing filters.
What to do next
Our Grok pages stay safe for work on purpose. If a scene cannot be said calmly, it does not belong in this library. For other errors, use the contact page.