It's a subset of one. "Deepfake" is the umbrella term for any AI-swapped or synthesized likeness; an AI nude video is the explicit subset — either a photo animated into fake nudity or a real face composited onto explicit footage. The concept map shows where each term sits.
In the strictest jurisdictions, yes — the UK criminalizes creating an intimate deepfake of a real person without consent, and even requesting one, with no distribution required. US state law is moving the same direction. The wider principle: law is migrating from punishing sharing to punishing creation. See the rules lesson.
Startlingly few — typically three to ten photographs, roughly what any public social-media profile contains. Identity encoders pretrained on millions of faces interpolate the missing angles. The teaching point: a normal online presence is already sufficient input.
Same engine, different mask. An undress tool masks the clothing region of a real photo and synthesizes beneath it. A face-swap tool masks the face region of existing footage and re-renders the target's face inside it. Both are diffusion inpainting; both produce identifiable imagery of a real person; the legal analysis is identical.
Assume yes — it's the safest default. Terms of service in this niche typically claim broad licenses over uploads, specify no retention limit, and name no accountable entity, while security research keeps finding their upload storage exposed. The upload is the irreversible step: it happens before any output exists and can't be undone.
Sometimes — seam flicker at the jawline, inconsistent earrings or teeth across frames, smearing when something crosses the face. But good outputs pass casual viewing, which is the point of the technology and the reason victims' audiences believe what they see. Detection is a triage skill, never proof; that's why the law doesn't require victims to prove fakeness.
Because enforcement hits their edges: registrars suspend domains, payment processors screen out the category, app stores remove clients. The backend is portable, so operators treat domains as disposable packaging — a pattern called domain churn. It's also why any link claiming to be "the official" version of one is unverifiable.
Four steps, in order: preserve evidence (screenshots, URLs, timestamps) before anything is removed; register the imagery with StopNCII.org (adults) or NCMEC Take It Down (minors) so platforms block re-uploads; file platform reports plus a US TAKE IT DOWN notice — which carries a 48-hour removal right; then pursue the legal layer — police in the UK where creation is a crime, a DEFIANCE Act civil claim in the US. Full walkthrough in the rules lesson.
Different analysis entirely. Synthetic content depicting a genuinely fictional adult — no real person's face, body, or identity — falls under ordinary adult-content rules, not NCII law. The legal line is identifiability of a real person, plus consent; the moment a real likeness is involved, the three-question test from the rules page applies.
No — deliberately. Every lesson here stops at the level needed for understanding and protection: what the pipeline is, where the harm enters, what the rules are. There are no tool names, no links, and no operational detail that would function as instructions. The about page explains the classroom philosophy.