Black Friday

50% off a whole year of any plan

Get the deal
eroq

blog/tutorials·Nov 21, 2026·6 min·by the eroq team

Uncensored AI prompting, responsibly — what renders and the lines that never move

Uncensored on eroq: your fiction renders as written, under one written policy. The lines that never move, how they're enforced, how to write dark fiction.


On eroq, uncensored means the engines render the fiction you write — dark, violent or unsettling stories included — instead of refusing them, under one written acceptable-use policy that applies to everyone the same way. It never means anything goes: anything involving minors, real people without their consent and anything illegal is refused before it renders, and a refused request costs nothing.

That is the same idea Venice made popular for private, unrestricted AI: no provider filter second-guessing your story, a policy you can read, and engines that keep nothing upstream. This lesson covers what that means in practice, which engines it applies to, how the rules are enforced, and how to write fiction that renders the first time.

What "uncensored" covers here

Most engines on eroq belong to other companies — Seedance, Kling, Veo, Hailuo, Runway, Wan, Grok Imagine, Nano Banana, GPT Image, Seedream and the rest. Each applies its provider's own moderation: a scene the provider turns down is refused before it renders, returns content_blocked, and costs nothing.

eroq's own engines have no provider filter in between. They render what you write, within the acceptable-use policy:

  • Video: eroq Motion One and Motion Pro.
  • Images: Eroq Uncensored (photoreal), Eroq One (references and editing) and Anima (anime and illustration).
  • Replace: eroq Replace Pro, which puts your characters into a clip.

They are also the private ones: eroq runs them itself, they retain nothing upstream, and nothing is used for training. What they render stays private to your account — it is never published to the community or posted to a social platform.

The lines that never move

These hold on every engine, in every tool, for every account, whatever the plan:

  1. Minors. Any content involving minors — real, fictional, stylized or implied. Every character is 18 or older, without exception.
  2. Real people without consent. No deepfakes, no impersonation, no recognisable person who did not agree, and no cloned voice of someone who did not agree. Your own face and voice are fine, and so is someone who gave you their documented consent.
  3. Sexual violence, real or fictional.
  4. Anything illegal where you or your audience are, and anything that helps commit a crime.
  5. Harassment, threats and hate, and promoting terrorism, extremism or self-harm.
  6. Deception: fake testimonials, impersonating brands or officials, synthetic footage passed off as real.

And one more that people miss: trying to get around the checks is itself a violation. Rephrasing, misspelling, encoding or splitting a request to slip a banned subject through is treated as the subject itself. As the policy puts it, a blocked request is a warning, not a puzzle.

How it is enforced

  • Before anything runs. Every request that makes an image, a video, a voice line, music or a chat reply is screened before it runs and before any credit is spent: a word filter in several languages, then a model that reads what the request means. Pictures and clips sent with a sensitive request are read by a vision model, and characters and their sheets get the strictest check.
  • A refusal is a code, not a charge. The request returns content_blocked and costs nothing. Why my render was blocked explains how to read it.
  • After a render. Renders of the uncensored engines are rated by a vision model once they land.
  • Before anything goes public. Every community post, film and public voice is reviewed by a person first.
  • On the account. The policy is enforced on accounts, not by an engine changing its mind: a violation can mean a warning, removed content, a suspension, or a report to the authorities with the evidence kept.

Private mode does not change any of this. It keeps no file, but private requests are screened before they run like any other (Private mode explained).

The Content filter on the web

The web studio has one more switch, under your account's Settings: Content filter. It "keeps sexual content out of your image and video requests", and it is on for accounts created since 18 November 2026.

  • When a request runs into it, the studio asks "Turn off the content filter?" and relaunches your request once you confirm.
  • Only the account holder can turn it off, in Settings or from that question, by confirming they are 18 or older. You can turn it back on at any time.
  • It applies to image and video requests made in the web studio. It changes nothing in chat, characters or voice, nor in the policy above, which applies whether the filter is on or off.

The mobile app goes further and refuses sexual and graphic requests altogether.

How to write dark fiction that renders

The checks read what a request means, so the clearest prompt is also the one least likely to be misread.

  1. Frame it as fiction. Name the genre, the setting and the stakes: "a 1970s crime thriller", "a war film, the morning after the battle", "a gothic horror set in a flooded manor".
  2. Say how old people are, plainly. "A detective in her forties", "two dock workers in their thirties". The minor guard is deliberately broad, so a word that reads young can trip it even in an innocent prompt; a plain age removes the doubt.
  3. Invent your people. Original names and original faces — or your own characters, cast with @mentions. Never a real name, a celebrity lookalike or someone's photo without their consent.
  4. Put the intensity in the filmmaking. Light, framing, sound and aftermath carry a dark scene further than a list of details: a single bulb, a long silence, a door left open.
  5. Pick the engine for the scene. A moderated engine may decline intense fiction; for those scenes, use Motion One in Video or Eroq Uncensored in Image.

A few prompts in that spirit:

A tense interrogation room in a 1970s precinct, a weary detective in her forties across the table from a bruised suspect, a single hanging bulb, cigarette smoke curling under it, high-contrast film noir photography.

A war-torn street at dawn, a lone soldier in his thirties walking past burned-out cars, ash drifting like snow, slow tracking shot at shoulder height, distant artillery, muted grey palette.

A flooded cellar in a gothic manor, a woman in her fifties holding a lantern up to a door that should not be there, water rippling around her knees, something knocking from the other side, low-key lighting.

A revenge western: a scarred gunslinger in his sixties waits alone in a dusty saloon at high noon, flies buzzing, the swinging doors creak open, slow push-in on his eyes.

When an innocent prompt is blocked

It happens: the guard is broad on purpose. Read the code, find the word that could read as something it is not — often an age-ish word — and say what you meant plainly. If a third-party engine declined it, the scene may simply be beyond its provider's rules: render it on one of eroq's own engines instead. What not to do is hunt for a spelling that slips a banned subject past the check.

FAQ

Is eroq's uncensored AI free to try?

Yes. A new account starts with 40 free credits, no card needed. Eroq Uncensored and Motion One are on every account, pay-as-you-go included: an image costs 10 credits / image, and Motion One 29 credits / second.

Why did Seedance refuse my prompt when Motion One rendered it?

Seedance, like every third-party engine, applies its provider's own moderation on top of eroq's policy. Motion One is eroq's own engine and applies the policy alone. Both refuse anything the policy bans.

Can I publish what I make on an uncensored engine?

No. Renders of the uncensored engines stay private to your account: they are never accepted on the community or posted to a social platform from eroq.

Does Private mode skip the checks?

No. Private mode keeps no file, but every request is screened before it runs, private or not.

Where can I read the full rules?

In the acceptable-use policy, and in short in the help center: The content policy and I got content_blocked. For the wider picture, see What unrestricted AI means.

Start in Image with Eroq Uncensored, or in Video with Motion One.

Render it in Image

Everything above happens in the studio: same engines, same settings, rendered on your click.

40 free credits · no card · private mode on every account

Prompts from this article

A tense interrogation room in a 1970s precinct, a weary detective in her forties across the table from a bruised suspect, a single hanging bulb, cigarette smoke curling under it, high-contrast film noir photography.

Privacy

Private and unrestricted

  • →No filter on eroq's own engines: one written policy instead.
  • →They keep nothing, and nothing trains on your work.
  • →Private mode keeps no file at all.