Licensing video for AI training: every rights holder inside one clip
One clip can hold the camera operator's copyright, performers' releases, bystanders' faces, licensed music and trademarks. Each layer has to be cleared separately.
Rights holders inside one clip
- Footage Staff work or contractor assignment.
- People on camera Releases and biometric consent.
- Music & marks Audio beds, logos, artwork.
Clearance inventory per clip
- Who owns the footage: staff work-for-hire or a contractor whose agreement may not assign rights.
- Talent and model releases: do they mention machine learning or 'any media now known or later developed'?
- Identifiable bystanders: face geometry can be a biometric identifier under laws like Illinois BIPA.
- Music and audio beds, often licensed for a single production only.
- Logos, artwork and private interiors.
Footage we won't help you offer
Security and surveillance video, bodycam footage and anything involving minors without verified parental consent. Collection purpose and consent make these a poor fit for licensing.
Metadata that raises usefulness
- Resolution, frame rate, codec, duration.
- Scene and action labels.
- Capture conditions: lighting, indoor/outdoor, camera motion.
- Per-clip release status.
How this affects value
Per-clip release status and capture metadata are what turn footage into a licensable dataset.
See what shapes a dataset's value