Rules and limits
NSFW AI image generation: what is allowed, what is blocked
Fictional adult content is allowed on Vellria: adult subject matter is not what the check looks for, and nothing sits between an adult prompt and the render. Two categories are refused with zero tolerance and close the account: any depiction of a person who is not of age, and content depicting a real, recognisable person without their consent, public figures included.
Where the line is
The rules are short. Fictional adult content is permitted, which means invented people in invented scenes, and adult subject matter on its own is not what gets a prompt stopped here. Against that, any depiction of a person who is not of age is prohibited without exception, as is any material relating to the sexual abuse of children. Those are the categories the check exists for, and they are refused outright rather than queued for someone to review.
The second prohibition is about real people. Generating content that depicts a real, recognisable person without their consent is prohibited, and the terms name public figures explicitly so there is no gap to argue over. The permitted case is invention: an adult who does not correspond to anyone who exists. Content that breaches applicable law, infringes the rights of third parties, or serves a fraudulent purpose is prohibited on the same footing, and you alone are responsible for the lawfulness of what you generate.
The check runs on your prompt before the run starts, and it is built to survive being written around. Accents, character substitutions, look-alike letters and invisible characters are normalised before the rules are applied, so a spelling trick reaches the same answer as the plain wording. This is the one thing on the platform that reads your prompt and can stop it, and it is looking for a category, not for a tone.
Models
| Variant | What it does | Credits |
|---|---|---|
| Vellria Flash | The fastest and cheapest image model. For drafts, variations and quick tests. | 1 credit |
| Seedream 5 Pro Uncensored | The flagship image model, focused on detail and composition. 2K output only. | 9 credits |
| Seedream 5 Pro Uncensored Edit | The flagship editing model. Takes up to ten references and preserves identity and lighting. | 9 credits |
| Qwen 3 Pro Uncensored | Strong on long, complex prompts; stays faithful to scene composition and text. | 6 credits |
| Qwen 3 Pro Uncensored Edit | Edits using the images you provide as references, preserving character and scene. | 6 credits |
| Seedream 5 Lite Uncensored | Fast, cheap Seedream. Reaches 4K and is good at typography and poster work. | 4 credits |
| Seedream 5 Lite Uncensored Edit | Takes up to fourteen references and redraws the images you provide. | 4 credits |
What a breach costs
The consequence is written down rather than left to discretion. A breach of the content rules closes the account without notice and, where necessary, is reported to the competent authorities. The terms describe no warning step in front of that, which is the reason to read the line rather than to discover it.
One clause is worth singling out because it is the only place in the published policies where money already in an account is explicitly forfeit: if a breach of the age requirement is identified, the account is closed without notice and unused credits are not refunded. Everything else about refunds is ordinary and predictable. That is not.
The age declaration and what it commits you to
Before the site opens you are asked to confirm that you are over 18, and continuing is that declaration. It is a declaration and not an identity check. Nothing is verified against a document, and the confirmation is remembered in a cookie you can clear from your browser whenever you like.
That cookie and the sign-in cookie are the only ones the site sets, and neither is used for advertising, tracking or profiling. What the declaration commits you to sits in the terms: the service is available only to people aged 18 or over, and the clause above is what applies if that turns out to be untrue.
Who can see what you generate, and for how long
Generations belong to the account that ran them and are reachable only from it. Nothing is published anywhere, and there is no public feed for your work to surface in. The privacy policy states the exception plainly, and it is better read here than discovered later: the administrator operating the service can access generation records, submitted prompts and generated outputs, for the purpose of monitoring abuse and resolving account and payment issues. Administrator actions on credits and accounts are recorded.
Your prompt does not stay in one place either, and that is inherent to the service rather than a policy choice. The prompt text and the addresses of any reference images are passed to the third-party provider that performs the generation. The crypto top-up runs through a third-party payment provider; card details and wallet keys never reach us. If the account itself is the concern, one can be opened without an email address at all.
Nothing is held indefinitely. Generated files and uploaded references are removed once their window is up, while the billing record of what you spent stays; the page on how generation works gives the window and how it is reported.
Money, when a rule is involved
A run that fails on the provider's side refunds itself, which covers the ordinary accident and needs nothing from you. What follows is the other kind of question: money you put in and then want back.
Top-ups are made in cryptocurrency, and an on-chain transaction cannot be reversed, so a completed payment cannot be refunded through the payment provider. Once credits are on the account you can submit a request for a refund of the unused ones from the Support page in your console. Requests are assessed individually; where a refund is made, it is made in the cryptocurrency used for the payment and at that day's rate. A top-up you have not paid yet can simply be cancelled from the console, which is the cheapest moment to change your mind.
Frequently asked questions
Is adult content actually allowed here, or filtered like everywhere else?
It is allowed. Adult subject matter is not what the check is looking for, and there is no separate tier or approval step for it. What gets refused is a narrow and fixed set: anyone who is not of age, and real people in non-consensual content. Everything the terms permit runs on the same catalog and the same API as everything else.
Can I generate images of a real person?
Not without their consent. Depicting a real, recognisable person without consent is prohibited, and the terms include public figures in that. The permitted case is fiction: an adult who does not correspond to anyone who exists. Uploading a reference does not change the answer, because the rule is about the output rather than the input.
Does anyone look at my prompts and images?
The privacy policy states that the administrator operating the service can access generation records, submitted prompts and generated outputs, for the purpose of monitoring abuse and resolving account and payment issues. Administrator actions on credits and accounts are recorded. Otherwise your outputs are reachable only from the account that made them and are not published anywhere.
What happens if I break the rules?
The terms state that a breach of the content rules closes the account without notice and, where necessary, is reported to the competent authorities. For a breach of the age requirement, they add that unused credits are not refunded. No warning step is described anywhere in the document, so treat the rules as a line rather than as an allowance for a first offence.
Can I get a refund on credits I have not used?
You can ask. Requests go through [the Support page in your console](/panel/support) and are assessed individually; a refund that is granted is paid in the cryptocurrency used for the payment, at that day's rate. A completed on-chain payment cannot be reversed through the payment provider, which is why the request route exists. A top-up you have not paid yet can be cancelled outright.