From Deepfakes to Fictional AI Characters: A Technology That Split in Two
"Deepfake" and "AI-generated character" get used almost interchangeably in casual conversation, but they describe two different technical approaches with two very different ethical track records. Understanding where the term actually came from — and why the industry and researchers increasingly draw a hard line between the two — explains why a platform generating wholly new, fictional characters isn't the same category of tool at all.
Generate a fictional characterWhere the term actually came from
The word surfaced around 2017 on an online forum, describing a face-swap technique that grafted one real person's face onto existing video or photos of someone else, without that person's knowledge or consent. The technique itself traces back further — face-swap compositing existed in film production for years before — but the term "deepfake" specifically became attached to the non-consensual, real-identity-swap use case from the start.
Why the term became toxic
The following years saw the technique used for political disinformation and for generating non-consensual intimate imagery of real, identifiable people, prompting a wave of legislation. A growing number of jurisdictions now specifically criminalize creating synthetic intimate imagery of a real person without consent, and platforms built around that use case face demonstrated search-engine demotion and de-indexing as a category.
A different technical lineage: generating someone new
Text-to-image and text-to-video diffusion models solve a structurally different problem: instead of taking an existing real person's likeness and grafting it somewhere, they generate a person who has never existed from a text description or a fictional-character reference, with no real identity being copied at any point in the pipeline. Researchers studying this space — including work from MIT Media Lab on AI-generated characters — have specifically framed responsible practice around three principles: permission from anyone whose likeness is used, disclosure that content is AI-generated, and content that doesn't defame, deceive, or harass — with fictional characters, not real people, as the baseline acceptable use case.
Where a platform like Uncutly sits in that split
Uncutly's generation pipeline is built on the fictional-character lineage, not the face-swap lineage: every result depicts a new, AI-original character, and the content policy independently bans generating anything based on a real, identifiable person's likeness without consent — regardless of what the underlying models are technically capable of. That's a deliberate architectural and policy choice, not a marketing distinction.
FAQ
Is Uncutly a deepfake generator?
No. Deepfake technology specifically grafts a real, identifiable person's likeness onto existing media without consent. Uncutly generates wholly new, fictional AI-original characters and bans real-person likeness use by policy.
Is all AI-generated synthetic media technically a "deepfake"?
No — the term is generally reserved for real-identity face-swap or voice-swap techniques applied to an existing real person, not for de novo generation of a fictional character that never existed.
Is fictional AI-character content treated the same as deepfakes legally?
Generally not, since no real, identifiable person is depicted — but regulation in this space is still evolving quickly, so it's worth checking the specific rules in your jurisdiction rather than assuming.
Why does this distinction matter practically?
It explains why legitimate research and platforms increasingly separate "synthetic media of a real person" from "de novo generative content" as different categories with different ethical and legal treatment — the technology and the harm profile are genuinely different, not just the branding.
Popular NSFW templates
