If you own a video library, whether that's a production company's archive, a channel's back catalog, a broadcaster's footage or years of your own shoots, AI companies will pay to train their models on it. Buyers pay for footage they can't get from the open web, and they pay every time it's licensed. Read on to find out how a video deal works, what labs look for, what has to be cleared first, what you can earn, and how to get started.
You license your videos, you don't sell them
The search term is "sell," and the deal is a license. You keep ownership of your footage, you decide what's included, and the buyer gets permission to train on it under agreed terms. A non-exclusive license can go to many buyers, so the same library earns more than once. Our guide to what a data license is and how it works covers the terms every license sets, and our guide to exclusive vs non-exclusive data licenses covers why one deal rarely has to be the only one.
Who buys video for AI training
Video generation companies need large volumes of clean, high-quality footage so their models produce video that looks real. World model teams need footage that teaches physics, how things move, collide and stay put. Labs can't get that from scraping video platforms, because web video is compressed, watermarked and shaky, and because a single clip can carry the footage copyright, the music, the performers, identifiable people and visible brands all at once. Licensed footage with clear rights is the way around both problems. Our guide to licensed video data for AI training covers what the labs buy and why, from the buyer's side.
What AI companies look for in your footage
| What they look for | What it means for your library |
|---|---|
| High resolution, 1080p or 4K preferred | Your masters and original exports, since the compressed upload loses detail |
| Well-framed, clear subjects | Footage where the model can see what's happening |
| Diverse scenes and scenarios | A varied library earns more than hours of the same shot |
| Professional or well-shot footage | Stable, lit and in focus |
| Clear audio | The original audio track, where you own it |
| Content you fully own or have rights to | Your shoots, your productions, and anything your agreements cover for AI training |
Two kinds of footage are in especially high demand right now. Multi-camera video teaches models to keep a scene consistent across viewpoints, and high-end animation supplies the polished motion that generation models are judged against. Buyers license across the categories you'd expect, like scripted content, documentary and reality, sports, interviews and talking heads, nature and wildlife, aerial and drone, advertising, animation and new media, so a library doesn't have to fit one genre.
What has to be cleared first
The value of your footage depends on how clean the rights are. Before anything is licensed, you decide what's in and what's out:
- Footage you don't fully own, like client work where the client holds the copyright.
- Music you licensed from someone else, which stays out unless the license covers AI training.
- Clips covered by performer or talent agreements that don't allow this use.
- Anything you received under a confidentiality agreement.
This isn't legal advice, and if your library has layered rights, your counsel should look at the agreements before you decide. Our guide to how AI data licensing deals work covers the terms that get negotiated and the gotchas.
What you earn
You earn a royalty every time your content is licensed, with no hidden fees. There's no public rate card, because the price depends on how many hours you hold, the resolution and quality, how varied and unusual the footage is, which categories buyers are chasing, and how clean the rights are. The one audited reference point in the market is Shutterstock, which earned 104 million dollars in 2023 from licensing deals with OpenAI, Meta, Google, Amazon and Apple, and that was a catalog built over decades. Your library is priced on its own merits, and the quickest way to a real number is to start the process below.
How the process runs
Licensing through us runs in four steps:
- Tell us about your content. You sign the partnership agreement and answer a short survey about what you own.
- Upload your content. Drag and drop, cloud transfer, or ship physical media. Processing takes one to two weeks.
- We prepare your content. Our pipeline clears, processes and annotates every clip for AI training.
- Get paid. We license to top AI labs on your behalf, and every use earns you a royalty.
Licensing is private and direct to the labs under strict NDAs, with no public listings, and you track everything in your own dashboard.
Direct deal or marketplace
The labs do sign direct deals, and the ones that get reported are with brand-name catalogs that have legal teams and data teams. A library nobody has heard of rarely gets a direct conversation, because the labs don't publish a specification for what they want. A marketplace signs one agreement with you and licenses onward to many buyers, so you never negotiate with a lab yourself. Our guide to who buys company data for AI training covers the kinds of buyer and how you reach each one.
Where to start
The quickest way to find out what your library could earn is to start licensing on our content owners page. You tell us what you own, we tell you what buyers are looking for, and you decide what to include.
Where Troveo fits
Troveo is a data licensing marketplace. We work with more than 40 active buyers and we've paid rights holders more than 20 million dollars. You sign one agreement, you keep full ownership, we handle the processing and every buyer conversation, and you earn a royalty every time your content is licensed. Start licensing your library, or talk to us about what you hold.