The stream could not be loaded. Check that the URL is correct and that the server allows cross-origin requests (CORS).
Vragenuur, 13 October 2026
Audio
Subtitles
About this demo
What you are watching
Three audio tracks
The original Dutch audio plus English and French dubbed tracks (AI-generated, for demonstration). The video stays the same; only the sound changes.
Subtitles in Dutch and French
Delivered as WebVTT subtitle tracks. Switch them on or off at any moment, with any audio track.
Dubbing with Arbor Signum
Reach viewers in their own language, and show it comes from the source
The dubbed voices sound natural, yet the lips still move in Dutch. Viewers notice that, and some will wonder whether what they hear is real. With Arbor Signum, Content Credentials (C2PA), every viewer can check where the video and its audio tracks come from.
Accessibility with dubbing
People follow the debate in English or French, with subtitles when they need them. Language is no longer a reason to stay away from the session.
Provenance with Arbor Signum
A signed C2PA manifest names the publisher, records the history of the video and states that the dubbed audio is AI-generated.
Together
The institution stays the official source. The dubbing is made in its own Arbor platform and published by the institution itself. A copy that lacks these credentials, or fails the check, can be told apart from the official version.
EU AI Act · Article 50
Transparency about AI-generated audio, built into the player
Since 2 August 2026, Article 50 of the EU AI Act requires transparency about AI-generated content. For dubbing, two duties matter. Arbor Signum and the player support you in meeting both.
Article 50(2)A machine-readable mark
AI-generated audio must carry a machine-readable mark, so it can be detected as AI-generated. The signed C2PA manifest from Arbor Signum records that the English and French tracks are AI-generated.
The EU code of practice expects a second layer: an imperceptible watermark. Arbor Signum watermarks the video today; audio watermarking is on our roadmap.
Article 50(4)A visible label for viewers
A dubbed voice that resembles the real speaker can count as a deep fake under the AI Act, so viewers must be told. The player shows an “AI-dubbed audio” label whenever a dubbed track plays, with the details in the Content Credentials.
Arbor provides the measures; the institution, as publisher, remains responsible for its own compliance. For AI systems already on the market before 2 August 2026, the marking duty in Article 50(2) applies from 2 December 2026. We are glad to go through this with your legal team.
AI helps people listen. The source still matters.
See the Arbor Signum demo →Dubbing workflow
Arbor sends one package, ElevenLabs returns one track per language
What do you think?
This demo shows an idea, not a product. Before we decide anything, we would like to hear from the people who would use it.
- Would your viewers use dubbed audio, or are subtitles enough?
- Which languages would matter most to you?
- Would you use it for live sessions, recordings, or both?
- Would Content Credentials on the dubbed audio help your viewers trust it?
- How do you plan to meet the AI Act labelling rules for AI-generated audio and video?
Share your thoughts with your Arbor Media contact.