“Text Check Failed”: When a Character or Scene Description Is Blocked Before the Image Is Made

Open the failed card, read the reason line, and edit the description right there before pressing regenerate.
Where the block shows
In Step 2 the card turns Failed with a short reason; hover it to see the provider’s original message, which names the text check (or, less often, the image check). Nothing was generated and the credits come back: the provider rejects before it draws. Open the card, edit the description, regenerate. The wider rules and what repeated blocks mean: Content Rules and Moderation.

The kinds of words that trigger it
Explicit and gory: nudity, sexual acts, wounds described in detail, blood as a feature. Real people: a celebrity’s name, “looks like” a named actor, a politician. Brands and IP: logos, product names, characters from known films, games or comics. Weapons and drugs named as the point of the image. Words that read wrong out of context: a “shot”, a “kill”, a “strip” of cloth, a child in a bath — the check reads words, not stories. SceneMixer checks Chinese and English alike.
How to rewrite
Describe what the camera sees, in neutral words. A soldier “covered in blood after the massacre” becomes “dust on the uniform, a bandaged left hand”. “Looks like [actor]” becomes the face itself: jaw, brow, eyes, hair. A brand becomes the object: “a red energy drink can, no label”. A named franchise character becomes an original design with its own anchor feature. Keep the identifying features — scars, glasses, a hairstyle — and drop the adjectives that judge. The same discipline makes better sheets anyway: Write a Description That Works.
When the image, not the text, is blocked
Sometimes the description passes and the generated image fails the check — the message names the image. The cause is still usually the description: it asked for something the model drew too literally. Soften the same words; if it was a photo you uploaded, the photo itself may contain a real person or a logo. Use Your Own Photo.
IP suspected
A separate check flags likeness to known franchises and brands. It is triggered by names, by iconic outfits and by realistic styles that land on a famous face. Replace the reference with an original design and change one anchor feature; switching to a stylised visual style often clears it. Blocked by Moderation: What to Change.
| Blocked wording | Neutral rewrite |
|---|---|
| covered in blood after the fight | dust on the jacket, a bandaged left hand |
| looks exactly like [famous actor] | square jaw, heavy brow, short grey hair, deep-set eyes |
| holding a [brand] phone | holding a black smartphone, no logo |
| a [franchise] warrior in the iconic armour | a warrior in dark segmented armour with a single red shoulder plate |
| a seductive nurse | a nurse in a plain white uniform, hair tied back |
FAQ
Was I charged for the blocked sheet?
No. The provider refused before generating; SceneMixer returned the hold to your balance.
Nothing in my description is offensive. Why was it blocked?
The check reads words, not intent. A single word that also has a violent or sexual meaning can trigger it; rewrite that word.
Does a block count against my account?
Not on its own. Repeated blocks in a short time make SceneMixer pause video generation; see the content rules page.
Can I ask support to allow my description?
No; the check runs at the provider. Rewriting is the only way through.
Is the check stricter for scenes than for characters?
Same rules. Scenes trip it with gore, drugs and real locations described as crime scenes.
Updated November 15, 2026