Acceptable Use Policy & Content Standards

Scope

These standards form part of our Terms of Service. They apply to prompts, uploaded photographs and reference material, generated images, videos and audio, shared links, and use of the image studio. They apply regardless of the selected model. The service is for adults aged 18 or older.

Six prohibited content categories

1. Pornography / NSFW: sexually explicit material, sexual services, fetish content, sexualized nudity, “nudify” requests and non-consensual intimate imagery. We do not offer an adult-content exception.
2. Violence / gore: graphic injury, dismemberment, torture, animal cruelty, glorification of violent abuse, threats, or material that promotes terrorism or mass violence.
3. Hate speech: content that attacks, dehumanizes, threatens or encourages violence against people because of protected characteristics, including race, ethnicity, nationality, religion, disability, gender or sexual orientation.
4. Child-unsafe content / CSAM: any sexual exploitation or sexualization of a minor, child sexual abuse material, grooming, or depictions that place children in abusive or dangerous situations. This includes synthetic content and subjects presented as minors. There is no exception for fictional or AI-generated material.
5. Deepfakes / impersonation: face swaps between real people, “deep nude” transformations, fabricated intimate scenes, identity theft, false endorsements, and fabricated representations of a real person's speech or actions, including non-deceptive fictional clips. Do not impersonate celebrities, public figures, private individuals or organizations. Permission does not authorize these prohibited uses.
6. Copyright / trademark infringement: material you are not entitled to use, counterfeit branding, misleading use of logos, or outputs that infringe another party's copyright, trademark, privacy or publicity rights. A model's ability to generate something is not a license to use it.

Do not evade safeguards by changing models, rewriting blocked requests or creating additional accounts. Spam, malware, fraud, harassment, unlawful surveillance and instructions facilitating terrorism or weapons of mass destruction are also prohibited.

Uploaded material and AI disclosure

Real-person generation is paused, including consensual self-portraits, public figures and real people in the background. Use pets, objects or non-human scenes that you own or have permission to process. Do not upload real-person photographs or voices. Consent does not override this restriction or the prohibited categories above. Do not describe generated output as authentic footage or use it as evidence of real events. The generator identifies its output as AI-generated; keep that context when sharing.

See AI models for the image and video models supported by the service. We are an independent service, not an official product of those model providers.

How moderation currently works

• Before generation: server-side rules screen prompts, with Waffo Prompt Sift providing a semantic text check when enabled. Uploaded images and generation references also undergo the media checks described below. A block, review verdict or unavailable check stops a new request before generation and credit reservation.
• Before delivery: the platform calls DeepSeek V4.1 Flash through Kie, with reasoning effort set to low, separately from the generation request. Only a timeout or service outage triggers GPT-6 Luna through Kie under the same rules; a rejection or a request for human review never does. Images are reviewed directly. Video is decoded by Cloudflare Browser Run into samples at four frames per second; its actual audio is transcribed with Whisper large-v3-turbo on Cloudflare Workers AI. The review model checks the sampled frames together with the transcript against the six prohibited categories and the real-person restriction. A verified absence of an audio track is recorded as silence; an audio-decoding error is not treated as silence.
• Release controls: only an explicit allow verdict permits a new result to be archived to public media storage, displayed or downloaded. A rejected video is withheld and its reserved paid credits are returned. An unavailable video check keeps the same generated output pending for retry, without another generation charge. We retain upstream model safety filters as an additional layer.
• Public shares: a conversation must pass text and image review before a public share is enabled. Changing its messages invalidates the approval. Published conversations are scheduled for another review daily; failed or expired checks remove public access. These controls apply to application share pages; a previously disclosed direct media URL is not a revocable private link.
• Reports and manual review: reports enter our private support-ticket queue for an administrator to investigate. Child-safety and credible imminent-harm reports take priority. Automatic decisions can be appealed through the same channel.
• Limits: frame sampling is not examination of every video frame. Transcription and model classification can miss content, context or rights violations. An automated allow verdict is not a copyright clearance or a human approval. Do not redistribute unsafe material; report it.

We retain media hashes, model and policy identifiers, verdicts and task outcomes. Approved identical media may reuse an approval for up to 24 hours; scheduled public-share rescans perform a new review. We use these records for troubleshooting and review. Administrators can update the additional prompt blocklist in response to observed abuse. We do not promise round-the-clock staffing or a fixed resolution time.

Reporting and appeals

Use Report content to open a support ticket. Sign-in is required for this channel; no purchase is required. Select a clear category, give the content URL or task ID, explain the concern, and state whether you are a rights holder or an affected person. Do not attach illegal or explicit material; supply a reference and a written description instead.

An administrator reviews the report, requests clarification where needed, and records the decision in the same ticket. Report details are not published. Information may be shared with service providers or authorities when necessary to investigate abuse or meet legal obligations. You can request reconsideration by replying to the ticket with the decision reference and new information. This channel is not an emergency service.

Enforcement

Depending on severity, recurrence and applicable law, we may issue a warning, reject a request, remove or disable access to content, restrict generation, suspend or terminate an account, or revoke promotional or improperly obtained credits associated with abuse. Serious child-safety violations, credible threats and other unlawful conduct may be referred to the appropriate authorities. We do not promise automatic refunds for prohibited use; statutory consumer rights still apply.

Actions are assessed in context. An accidental false positive can be reported through the same ticket channel. We will update this page when our rules or moderation process materially change.