How is a video generated with artificial intelligence labelled?

The practical rule is simple: if the video is generated with artificial intelligence, tell your audience, visibly and in text. The European framework adds machine-readable marking requirements on top of that.

Short answer

State publicly that the video was generated with artificial intelligence, through a written disclosure next to or inside the video.
Regulation (EU) 2024/1689 on artificial intelligence sets out, in Article 50, transparency obligations for synthetic content — image, audio and video.
Providers of systems that generate such content must mark it in a format that can be detected automatically.
Those publishing material that imitates real people or events must indicate that it is artificially generated or manipulated.
The regulation applies in stages and has particular cases, so check the text in force and seek legal advice for sensitive cases.

What the European framework provides, briefly

The European artificial intelligence regulation — Regulation (EU) 2024/1689 — addresses synthetic-content transparency in Article 50.
The marking must be machine-detectable, for example through metadata or other technical signals embedded in the file.
Beyond the technical marking, the public must be clearly informed when the content imitates real people or events.
The text also provides for particular cases — artistic content, for instance — so the exact wording is worth reading at the source.
Application is staged, and implementation guidance is still developing at European level; do not assume deadlines from memory.

How to label it in practice

Put a short line in the video description: “Video presented by an AI avatar”.
Where the context calls for it, add a one-second frame at the start or end with the same disclosure.
Write the disclosure as text on the page hosting the video too, so it is readable without the player.
Keep the technical markings added by the platform — do not re-export the video through tools that strip them.
Document internally which script, avatar and voice were used for each published item.

Common mistakes

Presenting the avatar as a real person from the company, with an invented name and job title.
Putting the disclosure only in an image, or at the end of a long text where nobody reaches it.
Re-exporting the video through an editor that strips the metadata added at generation.
Assuming a gallery avatar needs no disclosure because it does not resemble anyone in particular.
Treating labelling as a reputational risk. A short disclosure does not spoil the video and it avoids later arguments.

How 4avatars helps

4avatars generates videos with synthetic-content technical markings included and gives you the transparency disclosure as a reusable brand-kit element — a closing frame, description text, or both. That keeps labelling consistent across all your material instead of depending on who publishes it. For sensitive cases, check the regulation text in force and seek legal advice — the platform gives you the tools, not the counsel.