AI labs already train on gameplay. ZENOS makes sure you get paid for it. We license, capture, and sell the data. You do nothing.
Non-exclusive · 50/50 net · capture-ready in 24h. You keep your rights.Two things already turn your games into training data, and none of it pays you.
Public gameplay is scraped and labelled for training. Lossy, indirect data, and still your IP, taken without payment or permission.
Labs build bots to play your game, then run them at scale to collect the data themselves. Richer than scraping, no licence, nothing paid back. One lab could generate 100,000 hours, worth $6M+ at market, and owe you nothing. ⚑
Licensing through ZENOS turns both into rights-cleared, paid supply you control. Set the scope, get paid as it sells, and pull any title at any time. And licensed data is what labs actually need: training on unlicensed game IP is a legal exposure they would rather not ship.
From licence to payout, ZENOS handles capture, processing, QA, labelling, and lab sales. Onboard a title in a day, on sale the same week, payouts as labs buy. You keep making games.
One non-exclusive licence. You keep your rights.
On your released build. No SDK, no source, no engineering.
Labs license the library.
Net revenue, split 50/50.
Other data buyers ask you to install an SDK, ship a special build, hand over source, or capture thousands of hours yourself. ZENOS asks for none of it. We capture on your released game.
We recover direct production costs, then split net revenue evenly. Your data isn't sold once and gone. Produce a minute once, license it to many labs.
Illustrative. Your 50% share at mid-band pricing, from bundle sales. ⚑
| Scale | Hours sold / month | Your share / year |
|---|---|---|
| Early traction | 100 | ~$108k |
| Established demand | 1,000 | ~$1.1M |
| Breakout title | 10,000 | ~$11M |
In a new market, a one-off buyout sets a price before anyone knows what the data is worth. Shared upside means you participate as demand grows.
We capture the original released build. No mods, no source, no assets. Nothing that could rebuild or redistribute your title.
Models train on diversity, so most purchases need 30 or more titles across genres, sometimes hundreds. The value isn't in any one game on its own. It's in the catalogue it joins, and in the work that makes that catalogue usable.
Labs buy breadth: dozens to hundreds of titles spanning genres, mechanics, and worlds. Any one game is a single slice of what a training run needs. ⚑
That breadth is only useful once every title is normalised and labelled to one common standard, so data from very different games trains together. That is real work, and even done for one title alone, a single game is not what labs buy. Inside the ZENOS catalogue it is handled across the whole library, and your game earns as part of a set labs can actually train on.
Free evaluation. No cost, no commitment, no work. We benchmark your titles and tell you what labs would pay. ⚑