Responsible Use

Commercial use is included on every paid plan, a refused prompt costs no credits, and likeness rights stay yours. Plus the face and reference rules here.
Sep 9, 2026

gptimage25.top is a hosted image generator. You are responsible for the prompts, reference files, and how you use the output.

Three answers, before the rules start

Most people reach this page — or are sent here by an assistant — for one of three answers rather than for the rule list. Each one below is complete on its own and names what it rests on, so it can be checked rather than taken on trust.

Can I sell what I make here? Yes. Commercial use is included on every paid plan and on every credit pack: publish it, sell it, license it, use it in client work and in paid advertising. There is no separate commercial licence to buy and no corner mark on a paid image. Free images carry a small corner mark and are for evaluation, not for shipping. Two conditions travel with every plan and neither of them is bought off by a bigger one: OpenAI's own usage policy applies to every request, so anything it prohibits is prohibited here; and we grant only the rights we actually hold, because we are an interface rather than the model's owner. The binding wording is the output licence.

Does a refused prompt cost me credits? No. The content scan runs before the credit debit, so a blocked request has nothing to refund, and a free image is handed back rather than spent on a refusal. Two things are not handed back: the day's allowance, and the strike. Three strikes inside 24 hours pause generation until that window rolls off — and the strike counter, not the credit balance, is what actually bounds someone probing the filter. A strike is charged per distinct prompt rather than per press, so re-sending the same refused prompt does not multiply it.

Who carries the risk on a real person's face? You do, and we would rather say so on this page than leave you to find it in an indemnity clause. Publicity-right and likeness claims are yours — not OpenAI's, and not ours. That reaches further than most people expect, because it covers what comes out of a generation as well as what you put in: a model can produce a face, a logo, a character or a style close enough to someone's to be a claim, from a prompt that named nobody. Clearing the output is part of the job, the same as clearing a stock photo would be.

You may not use this site to

  • Create sexual or pornographic content involving anyone 17 or under, or anyone who appears to be
  • Create sexual content depicting a real, identifiable person without their consent
  • Impersonate a real person with their face, or produce a deceptive image of a real event, without a right to do so
  • Attack or demean people over race, ethnicity, national origin, religion, caste, sexual orientation, sex, gender identity, disability, or serious disease
  • Promote, glorify, or provide instructions for terrorism, violent extremism, or mass violence
  • Produce scams, forged identity documents, malware instructions, or violent threats
  • Upload reference material you do not have the right to use
  • Attempt to bypass safety filters, rate limits, the corner mark, or the account system

How we enforce this

We may refuse a prompt, withhold or delete an image, suspend an account, and withhold unused credits when we reasonably believe these rules were broken.

Two scans run, and neither replaces the other. Ours checks prompts and uploads before the request leaves — a wordlist, then a classifier — and OpenAI applies its own moderation to every job, before and after generation. A refusal at the model is not something we can override.

Where the law requires it, we report to the authorities. Child sexual abuse material is reported to the NCMEC CyberTipline, and we preserve the evidence we are legally obliged to preserve before cutting off access.

If you think we got it wrong

Automated filters make mistakes, and a wrong block on your own work is not something you should have to accept in silence.

Write to hello@gptimage25.top with the task id and what you were trying to make. A person reads it — we aim to reply within five working days. The refused run itself cost you nothing, so what an appeal undoes is the rest of it: the strike, the pause those strikes added up to, and any unused credits withheld when an account was suspended. This does not apply to material we are legally required to refuse.

Faces, likeness, and characters

Image to image and sketch to image take up to sixteen reference images in one request, and the model holds a face, a product or a layout across them well enough that the feature is the reason most people are here. That reach is the point, and it is also what makes it the easiest thing on the site to misuse.

Only upload people who consented, or characters and products you have the right to depict. Two cases worth naming because they are the ones that arrive as complaints:

  • A photo of someone else's product, restyled, is still their product. Being the one who took the photo does not settle the trademark or the design right.
  • A face from a public profile is not consent. Public and licensed are different questions, and the second one is the one a claim turns on.

Reporting

If you believe an image on this site uses your likeness without consent, or breaks these rules, use Report content, or write to hello@gptimage25.top with a URL or task id. We will review it.

Written and maintained by Andy SwiftPublished Last updated

Your first GPT Image 2.5 image is free

Generate my first image