Plain-English explanation
Synthetic media is the general category that covers essentially all AI-generated or AI-modified content: text-to-image outputs, AI-generated video, cloned or synthesized voices, AI-written text, and AI-altered photos or footage. The term doesn't imply anything about whether a real person is involved: it's defined by how the content was made (AI generation or modification), not by what or who it depicts.
This is a meaningfully large category, and industry groups such as Partnership on AI have published frameworks specifically for the responsible creation, disclosure, and handling of synthetic media, treating it as a distinct area of practice given how broadly it now spans creative, commercial, and communication contexts.
Adult-AI use
An AI-generated companion character, a text-to-image render of an invented persona, and an AI-voiced roleplay response are all synthetic media. They are not deepfakes because they do not depict a real person's likeness. The distinction matters because "deepfake" carries specific non-consent implications that do not apply to activity built around wholly original characters.
Common uses
- AI-generated original characters used in companion apps, chatbots, and image/video generation.
- AI voice synthesis for narration, dubbing, or a companion's spoken responses.
- AI-written text: chatbot dialogue, generated stories, marketing copy.
- AI-based photo or video editing that goes beyond simple filters: object removal, style transfer, upscaling.
How it differs from deepfake
Synthetic media is the category; deepfake is a specific, narrower subset of it involving a real, identifiable person depicted doing or saying something they didn't. All deepfakes are synthetic media; most synthetic media is not a deepfake. An AI companion character with an invented name, face, and personality is synthetic media, not a deepfake, because there is no real person whose likeness or identity is being misrepresented.
Limitations and misconceptions
- "Synthetic media" and "deepfake" are sometimes used interchangeably in casual conversation and even in some news coverage, but they are not the same thing, and conflating them overstates how much AI-generated content involves real people.
- Synthetic media isn't inherently deceptive or harmful: the term is purely descriptive of how content was made, not a judgment about its purpose, quality, or honesty.
- Disclosure can use watermarking or Content Credentials, but labeling norms are an active area of platform policy and, in some places, law: they're not a settled or universal standard yet.