← All guides

When generation is blocked: safe rewrites that still work

Automated filters, human gallery review, and how not to burn credits on disallowed asks

Nanomelon shows a policy or failure message when a job cannot complete. That is not a bug and it is not a dare. Upstream models and our own public-gallery rules exist so the site stays usable for a general audience and so Public Images does not become an unmoderated dump. This page explains common block reasons, what happens to credits, and how to rewrite toward something you can actually finish—and, if you want, publish under the community guidelines.

Two different gates

Gate A — generation. The model or API refuses the request. You see an error in the UI. Credits are not deducted for failed or policy-blocked jobs.

Gate B — public display. The image may generate privately, but if you marked “Publicly visible,” staff must still approve it before it enters Public Images. Pending items stay in your account only. Rejected public requests do not appear in the grid.

Confusing the two gates causes needless anger: a private experiment that looks edgy might generate, then fail gallery approval. That is working as designed.

Common generation block themes (and safe redirects)

Copyrighted characters and brand mascots

Asking for a named franchise hero or logo usually fails. Rewrite to an original character with similar job and palette: “original fox-masked courier in a teal jacket,” not the trademarked name. Templates for character sheets live in the prompt library.

Real public figures

Celebrity likeness prompts are frequently filtered. Write an age band and traits instead: “middle-aged woman with short grey hair, linen blazer,” not a living politician or actor.

Sexual content and minors

Any sexual content involving anyone 17 or under is prohibited—generation and public display. Softening the wording to evade the filter is still a violation and can get the account reviewed. Keep child-present scenes ordinary and non-sexual (festival clothes, schoolyard play) or write clearly adult characters.

Violence and illegal how-tos

Detailed instructions for weapons, crime, or self-harm are out of scope. Mood horror with fictional creatures is different from a tutorial. When in doubt, drop the instructional language.

Medical and “miracle” claims

Prompts that ask for before/after disease cures, unapproved drug labels, or diagnostic imagery of real patients often trip both model filters and our public-gallery rules. For ecommerce lifestyle work, show products without inventing clinical claims—see the ecommerce workflow.

Rewrite patterns that usually clear Gate A

Think of a blocked prompt as a sentence with one poisoned noun. Keep the useful structure (pose, place, light, medium) and swap only the noun.

• Brand → category: “Nike sneaker on marble” → “white running shoe with teal stripe on marble, no logos.”
• Celebrity → traits: “[Famous actor] in a trench coat” → “middle-aged man with salt-and-pepper beard, trench coat, rainy street, photograph.”
• Franchise hero → original job: “Spider-Man on a rooftop” → “original masked acrobat in a red suit with a geometric emblem, rooftop at dusk, comic illustration.”
• Suggestive teen cue → clearly adult or non-sexual child scene: remove sexual language entirely; if a child remains, keep festival/school contexts only.

Then set quantity to 1 and regenerate once. If it still fails, the problem is usually still a proper noun or a sexual cue you missed—not “the model hates you.”

What to do after a block

1. Do not spam the same prompt with tiny synonym changes. That wastes your time even when credits refund.
2. Remove the disallowed noun (name, brand, sexual minor cue).
3. Replace it with an original subject and a concrete place + light (see the five-part brief).
4. Retry once at quantity 1.
5. If you only needed a private mood board, turn public visibility off so you are not waiting on Gate B.

Private success, public rejection

Sometimes Gate A passes and Gate B fails. Examples: a borderline costume that looks fine privately but is too close to a trademarked mascot for the public grid; a prompt that includes someone else’s phone number; a near-duplicate spam upload. Staff can reject public listing without deleting your private file. Fix the prompt or keep the work private. Arguing that “the model already drew it” does not override gallery policy.

If you plan to publish, read the guidelines before you toggle Publicly visible. Approval is not instant; pending items never appear in the public API list.

Credits and honesty

Failed and policy-blocked generations do not take credits. Successful generations you dislike still do. That is why quantity 1 is the default for learning. Details: models and credits. Raising quantity on a prompt that already failed once is how people burn a session without learning anything—refunds apply to failures, not to five near-identical successes you hate.

Reporting bad public content

If something slipped into Public Images that should not be there, email support@flyingmob.com with the URL. We review reports as described in the community guidelines. Do not use the report channel to argue about a private job that was correctly blocked.

Include the gallery URL or a screenshot of the card, the reason (hate, sexual minors, infringement, scam), and a contact address. We aim to act within two business days on clear violations.

Why we publish this

Transparency about filters is part of trust. AdSense and users both need to see that UGC is not a free-for-all. Automated refusal plus human approval for public works is the mechanism; this article is the explanation in plain language. Paired reading: anime vs realistic (medium-first steering reduces “weird hybrid” retries that look like filter fishing) and the template library (skeletons that stay on the safe side of proper nouns).