Synthetic Media
Any media content — audio, images, video, or text — that was generated, modified, or substantially enhanced using AI models rather than captured directly from a camera, microphone, or human author.
"Synthetic media" is the broad, neutral umbrella term for anything made by a generative AI model. It covers:
- AI-generated images (text-to-image, upscaling, inpainting).
- AI-generated or -edited audio (TTS, voice cloning, voice conversion).
- AI-generated or -modified video (face swaps, lip-sync, reenactment, fully-synthetic clips).
- AI-generated text (articles, social posts, code, summaries).
vs. "Deepfake"
The two terms overlap but aren't interchangeable:
- Synthetic media is neutral. A magazine cover illustration generated by Midjourney is synthetic media but not a deepfake. A voice actor's TTS narration for an audiobook is synthetic media but not a deepfake.
- Deepfake implies intent to deceive. The same synthetic output becomes a deepfake when used to impersonate, mislead, or harm.
Regulators increasingly use "synthetic media" for the broader category because it captures legitimate uses (accessibility, localization, content creation) without prejudging intent.
Why the distinction matters for policy
Most new legislation — the EU AI Act, China's 2023 labeling mandate, various US state laws — regulates synthetic media at the labeling level (disclose that it's AI) and deepfakes at the criminal level (non-consensual imagery, election interference). See our country-by-country regulation guide.