EU deepfake label on a recorded video: disclaimer and end credits

For non-real-time video, the EU Code of Practice as summarised by Tech Policy Press calls for an opening disclaimer, a persistent icon and end credits. You can author the disclaimer and the credits as timed overlay text on a finished Sume clip with POST /v1/video-captions and cues, which burns text without speech-to-text. This is a workflow, not legal advice.

What the page says by format

The article says disclosure must happen "at the latest at the moment of first exposure". It lists real-time video as a persistent, non-intrusive icon with an opening disclaimer, non-real-time video as opening disclaimers plus a persistent icon plus end credits, images as a clearly visible fixed icon, audio-only as audible disclaimers repeated for longer content, and multimodal content as a visible icon that needs no user interaction.

It says providers must mark outputs in a machine-readable way, and deployers who create realistic synthetic content must disclose it.

Author the text with cues

Standalone video captions take a public HTTPS video_url and, for authored text, cues with text, start and end. Because no speech-to-text runs, it also works on a silent clip. This example puts a disclaimer in the first three seconds and credits in the last three of a 12-second avatar clip.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: disclosure-cues-001" \
  -d '{
    "video_url": "https://example.com/avatar-clip.mp4",
    "cues": [
      { "text": "AI-generated avatar", "start": 0, "end": 3 },
      { "text": "Made with AI. Not a real person.", "start": 9, "end": 12 }
    ]
  }'

What is left to you

Text cues cover the opening and the credits. The persistent icon is a still, so lay it over the video with Sume's timeline compose, which puts a still and a video in one frame. Machine-readable marking is a provider duty that I did not verify for Sume, so do not claim it.

Check your own obligations with counsel. See Tavus verbal and visual disclosure versus Sume caption cues for a related comparison.