Is a personal AI avatar video exempt from EU deepfake labels?

According to Tech Policy Press's summary of the EU Code of Practice, private individuals who generate AI content for personal, non-professional use are exempt from labeling obligations, and artistic, satirical, creative or fictional works need only minimal, non-intrusive disclosure. A video made for a business is a deployer's output and falls under the disclosure duty. This is a reading aid, not legal advice.

What the summary says

The article divides the duties between providers, who must make AI output machine-readable and detectable, and deployers, professional users who must tell people when realistic synthetic content is artificial. It names one exception: private individuals using AI for personal, non-professional purposes. For creative work it describes a lighter touch, a disclosure designed not to interfere with the work's integrity or enjoyment.

It also says disclosure must come at the latest at first exposure.

Sorting your own avatar videos

A talking-avatar clip falls into one of three groups. A birthday greeting you send a friend is personal use. A short comedy sketch with an obviously fictional presenter is creative work. A product explainer, a sales update or a training video is professional use, so label it.

Sume's POST /v1/avatar-1.0/talking-video makes all three the same way, so the label is your decision, not the tool's. You can add burned-in captions inline (captions.enabled) or put authored text through POST /v1/video-captions with cues.

When in doubt, label

Labeling costs seconds of caption text and removes the argument. Put the line in the first seconds of the clip, and keep a copy of the script and the job id so you can show how it was made. See labeling an AI avatar Instagram account for a platform-side version.