1. Scope and responsibility for materials
This policy covers user uploads, available text and parameter inputs, operator reference materials, AI-generated images, and images stored or shared through Lopumi. Submit only materials you have the right to use, with necessary rights-holder permission and express authorization from people depicted. Authorization does not permit prohibited real-person deepfakes, sexual exploitation or other abuse. This service does not offer audio or video creation.
2. Pornography and NSFW — prohibited
Pornography, explicit sexual acts, sexual exploitation, sexualized nudity and content intended to undress real people are prohibited. NSFW here means pornographic and sexually explicit content unsuitable for public or workplace viewing. There is no exception allowing it after adult age verification.
3. Violence and gore — prohibited
Severe gore, dismemberment, torture, abuse, and content encouraging, glorifying or assisting violence, terrorism or extremism are prohibited, including actionable guidance for mass harm or weapon manufacture. We offer no restricted-gore generation option.
4. Hate and discrimination — prohibited
Dehumanization, hatred, persecution, threats or incitement to violence based on race, ethnicity, religion, gender, sexual orientation, disability or similar identity are prohibited.
5. Child safety — zero tolerance
Child sexual abuse and exploitation material (CSAM), any sexualization, grooming, solicitation or exploitation of minors, and related synthetic content are prohibited, whether real, fictional or AI-generated. Uncertain age, fictional characters or purported consent cannot bypass this prohibition. On discovering such risk, we prioritize access restrictions and investigation and cooperate with competent authorities as applicable.
6. Deepfakes, impersonation and NCII — prohibited
Real-person face-swap impersonation, fabricated speech or conduct, celebrity or public-figure impersonation, fabricated intimate interactions between real people, intimate-scene composites and undressing are prohibited regardless of consent. Non-consensual intimate imagery (NCII: private or sexualized imagery created, used or distributed without the depicted person’s consent) is prohibited. Permission to use a photograph does not authorize these prohibited functions or purposes.
7. Copyright, trademarks, privacy and likeness
Unauthorized infringing reproduction, misuse of trademarks, fabricated authorization or endorsement, and violations of privacy, likeness or other rights are prohibited. Rights holders may request review through the reporting email with the specific work, URL, explanation of rights and necessary evidence. False rights claims are prohibited.
8. Other abuse
Fraud, harassment, threats, unlawful purposes, safety-control evasion and presenting AI results as evidence of real events are prohibited. Altered wording, model switching, repeat requests or auxiliary materials must not be used to bypass restrictions.
9. Actual review mechanisms and limits
Before generation: operators configure product templates, preview references and controlled options; the server validates versions, roles and allowed parameters. Users cannot submit arbitrary free-form prompts. Optional AI optimization uses the designated main image and operator template to produce group prompts, with output-protocol and frozen technical-contract validation. These structural and permission checks are not comprehensive content-safety classification. Generation and delivery: we honor model-provider safety refusals; recognized safety refusals do not automatically switch routes to evade the restriction. Provider capabilities and detection have limits. The platform does not currently have independent, comprehensive safety filtering covering every request before generation and every image before display. Upload format/size checks, collage CLIP checks and review curation are not NSFW or CSAM safety moderation. Platform review: when risks are discovered or reports arrive, the operator investigates using tickets, task records and necessary materials, records decisions, and applies access/sharing restrictions or further measures. We do not promise human pre-approval of every image, perfect detection, an independent review team or 24/7 human staffing.
10. Proportionate enforcement and retention limits
Depending on risk, evidence and repetition, measures include notices or warnings, refusal of requests, generation or sharing restrictions, temporary access restrictions, removal after investigation, account suspension, or ending access for serious or repeated violations. Urgent risks may be restricted before review; suspected serious illegality is handled with competent authorities as applicable. Intake, reasons, decisions and outcomes are recorded. Emergency restrictions may apply to an account or sharing-link scope. Deletion is subject to preview references, in-flight tasks and legally required retention. We cannot promise instantaneous complete deletion of all files or retrieve copies already downloaded by others.
11. Credits, refunds and existing contracts
A report does not automatically zero credits. Fraudulently, abusively or improperly obtained promotional or reward credits may be revoked on evidence under the original rules. Lawfully purchased unused credits, frozen tasks, refunds and disputes remain subject to the original contract, actual delivery and applicable rules. We do not automatically confiscate all purchased credits or remove statutory rights. Explicit failure releases the related reservation; genuinely unknown results remain under original-request verification and are not blindly resubmitted because of a report or review.
12. Reporting information and privacy
Include the category, relevant URL, task or order reference, brief facts, necessary evidence of your rights and a reply email. Do not send passwords, verification codes, payment credentials, suspected child sexual abuse material or non-consensual intimate originals; provide location references and a safe written description instead. Report information is used with restricted access for investigation, handling and legal necessity. We may request more information and record the review, decision, reasons and outcome.
13. Response times and urgent matters
Reports, appeals and support use the single email marcni1998@gmail.com. Email is received at all times. A person checks it daily, 09:00–21:00 Beijing time (UTC+8). Ordinary reports receive a first reply within one business day of receipt. That reply explains intake, information needed and the expected handling plan; it is not a promise to resolve the case within that period. Business days are Monday–Friday, excluding Japanese public holidays, with dates determined in Beijing time (UTC+8). Daily mailbox checks also include weekends. Child sexual exploitation, non-consensual intimate imagery and credible threats to personal safety receive priority: once we become aware, we immediately take emergency measures proportionate to the risk and evidence and review the matter, without waiting in the ordinary queue. An always-receiving mailbox does not mean 24/7 human staffing. For complex cases we explain progress or reasons for delay. Appeals use the same email; a person explains review arrangements. Statutory complaint and remedy rights are not restricted.
14. Appeals, notices and policy relationship
A person affected may request reasons and appeal to the same email. A person receives the appeal and explains review arrangements; urgent restrictions may precede notice. No internal appeal deadline excludes statutory rights. This standalone AUP is stored and effective with the Terms version shown on this page. Historical versions and acceptance records remain; where required, new creation or purchase uses the existing on-demand confirmation. Original AI-effect, selection and delivery explanations are retained in the following sections.
AI effects and choices
Styles and prompts are operator-defined; only the product's permitted choices are accepted. Multiple materials jointly reference one collection and do not determine output quantity. AI may make structural, species or text errors and produce similar results. Illustrations do not guarantee replication, exclusive copyright or suitability for every use. Check relevant rights before publication or commercial use and retain required AI labels.
Series, vision and delivery
Enabled pet products accept 1–5 references and 1–4 grids containing 4/8/12/16 candidates. Each grid is an independent preview; purchase up to 10 distinct available finals. Watermarked previews and internal originals/crops differ from final deliverables. Optional scene adaptation analyzes only one designated main-image thumbnail, adjusts unsuitable activities, and adds the displayed processing time and credits. It is optional. Ready grids can be selected immediately; completed finals can be downloaded separately without waiting for every item or reviewing the service.