Video training data
Video training data is a licensed collection of footage, often with captions, audio and metadata, used to train video and multimodal models.
Last reviewed
What video training data contains
- Raw footage
- Captions and transcripts
- Scene and object labels
- Rights and release records
Who buys it and how it is priced
| Attribute | Value |
|---|---|
| Belongs to | multimodal training data |
| Main buyers | Labs building video generation and understanding models |
| Pricing | Commonly priced per hour of footage, with rates driven by exclusivity, resolution, rights clearance and uniqueness. |
Quality criteria and risks
Quality criteria
- Clear rights and releases
- Resolution and frame rate
- Caption accuracy
- Diversity of scenes
Risks
- Missing likeness or music rights
- Duplicated stock footage
- Face and biometric privacy
Providers that list video training data
Auction and pricing mechanisms that fit
- Posted-price sale: Fits catalogue datasets that can be sold many times, such as cloud data exchange listings.
- Revenue-share licensing: Fits long-term content licenses where value is uncertain up front and both sides want to share upside.
- English auction: Fits one-off, scarce assets where buyers can see each other, such as a unique archive of footage or a rare-language speech corpus.
Key terms
- data licensing
- Data licensing is the grant of rights to use a dataset for stated purposes, in return for payment or other terms.
- data provenance
- Data provenance is the documented history of where data came from, who created it and what rights attach to it.
- opt-out
- An opt-out is a signal or process by which a rights holder tells AI developers not to use their content for training.
Frequently asked questions
- How is video training data priced?
- Most deals are quoted per hour of footage, adjusted for exclusivity and rights. Few prices are public.