Permitted inputs
Image to Video accepts still photos and a text description only. Video to Prompt accepts one clip of up to 60 seconds, analyses it into text and deletes it; it never edits the clip or uses it to generate video, and it describes people without identifying them. Upload only clips you have the right to use. For Image to Video, upload photos of yourself, or obtain explicit permission from every adult pictured for this use. You must also have the necessary rights to the photos. Public availability is not authorization. The uploader must actively confirm these conditions; a checkbox is not proof of identity or permission. Original fictional adult characters must not impersonate real people.
1. Sexual content / NSFW
No pornography, explicit sexual acts, nudity, sexualized exploitation, undressing tools or intimate-image abuse. These restrictions apply even when a person consents.
2. Violence / gore
No graphic injury, dismemberment, torture, glorification of violence, threats, or encouragement or instructions for self-harm or harm to others.
3. Hate and targeted abuse
No hateful or dehumanizing attacks based on protected characteristics, extremist propaganda, harassment or targeted intimidation.
4. Child safety / CSAM
No child sexual abuse material, sexual exploitation, grooming, or sexualized depictions of anyone under 18, including synthetic content. This adult-only service also rejects or holds media when a depicted person’s adulthood is uncertain.
5. Deepfakes / impersonation
We do not offer face swap, and we do not edit or replace people in existing videos. A new clip of yourself, or of an adult who explicitly authorized it, in an ordinary scene is the intended use. Prohibited regardless of consent: impersonating another real person, celebrity or political figure; fabricated statements, confessions, news or evidence; fake endorsements or fraud; undressing; kissing, hugging or other intimate interactions between real people; and presenting a result as real footage.
6. Copyright / trademarks
No infringing characters, artwork, recordings, logos, trademarks or other protected material. Do not imply affiliation or endorsement. Hold the necessary licenses; uncertain rights or recognizable protected identities may cause a request to be held or rejected.
Moderation before generation and delivery
Files you only preview locally never reach a model. The generation pipeline requires prompt checks, a check of every uploaded photo and an output check before any download. Its self-built safety layer combines versioned prompt rules, Gemini 2.5 Flash Lite semantic/visual classification through fal, and fal image NSFW checks. Generated videos are sampled at 32 frames: this is not exhaustive frame-by-frame or speech review, and cannot prove identity or permission. An uncertain, incomplete, expired-policy or unavailable check holds or stops the task. Unapproved outputs remain private and unavailable. Prompt Sift can add a separate prompt check; it is not a replacement for output checks. We retain moderation decisions and provider references, with encrypted response evidence subject to our retention and deletion rules. There is no public user gallery.
Reports, enforcement and appeals
Email support@selfscene.ai or use the signed-in report form. Include the task ID or page URL, the concern and any relevant rights information; do not send illegal imagery. Reporter identities are kept confidential, and every report is reviewed objectively. Response times from receipt: severe reports (child safety, non-consensual intimate imagery, terrorism or credible threats) get an initial response within 2 hours and are resolved within 24 hours; high-risk reports (impersonation or deepfakes of a real person, sexual content, hate, fraud) within 24 hours and 3 business days; medium-risk reports (copyright, trademark and other policy concerns) within 3 and 7 business days; other reports within 5 and 15 business days. While a report is investigated we may restrict access to the affected content; we may block tasks, delete content or suspend accounts, and we email the reporter the outcome. If more time is needed we tell the reporter why. You may appeal a decision within 30 days by replying with the reference and supporting information; appeals are reviewed within 10 business days, where possible by someone who did not make the original decision. Review does not automatically approve or publish content. Safety rules are versioned and have a mandatory review expiry; stale rules stop new moderation until reviewed.
Audio and copyrighted material
The model generates its own soundtrack for each clip. The service does not accept audio uploads, clone voices or provide licensed music. Do not prompt for a real person’s voice or for copyrighted songs, and do not imply that a result includes licensed music.