Whether posting an AI face-swap video online is illegal has less to do with "using face-swap technology" than with three things: whose face you used, whether they agreed, and whether the video looks real. Making a funny video of your family and sharing it only in the family group chat is basically safe; posting it on a short-video platform for traffic or using it commercially for sales crosses very different risk lines — step on one and you're infringing.
Check the scenario table first
| Scenario | Verdict | Risk |
|---|---|---|
| A funny face-swap video of your family, shared only in the family group chat or DMs | Generally fine | Low |
| Posted on short-video platforms or social media for views | You must have the person's consent; also add the platform's AI-generated label | High |
| Commercial sales, brand promotion, or swapping in a celebrity's face | Extremely likely to infringe; the platform may take it down and suspend the account | Very high |
The three boundaries
First: consent for the portrait right
China's Civil Code states that every natural person has a portrait right: no one may make, use, or publish someone's portrait without their consent. The person whose face you swapped in is the portrait-rights holder. Even if you used a publicly available photo and didn't earn a cent, swapping their face and publishing it publicly requires their consent. "Everyone is doing it" is not a defense.
Second: no uglification or defamation
Even with consent, you may not make the person look ugly, defiled, or place them in fabricated scenarios. The Civil Code is explicit: no organization or individual may infringe another's portrait right by uglifying, defiling, or forging with information technology. Face-swapping a public figure into an indecent clip or turning a teacher into an insulting meme is directly blocked by this rule.
Third: AI-generated content must be labeled
The Measures for the Labeling of AI-Generated and Synthesized Content (《人工智能生成合成内容标识办法》), in effect since September 1, 2025, require AI-generated content to carry explicit labels (visible to viewers) and implicit labels (embedded in file metadata). No organization or individual may maliciously delete, tamper with, forge, or conceal these labels. In practice: use the platform's built-in "AI-generated" label when posting, and say "AI face-swap" in the caption — don't use misleading wording like "real footage" or "it's really them."
Two steps to stay safe
- Get consent first. That goes for relatives and friends too: consent to share in the family group is not consent to publish publicly.
- Add the label. Put "AI-generated face swap" in the title, comments, or a pinned note — don't let viewers mistake it for real footage.
Common misconceptions: "no profit, no infringement" — wrong, profit is not a required element; "it's just for fun" excuses everything — wrong, a fun disclaimer can't replace the person's consent; "a blurry watermark counts as labeling" — wrong, deleting or faking the implicit label is itself a violation.
