AI that finally gets the plot.
Kinora reads film, TV and sport the way people do: who wants what, why it matters, how it feels. That turns every screen into something you can ask, reshape and shop, and every ad slot into a moment that fits.
By 2030, every story is something you can talk to, reshape and shop.
Three ways people will watch. Each card opens it in the demo above.
Big AI can't build the rails for this future.
Media is the largest untapped river of attention
People spend around four hours a day with TV, streaming, sport and creator video. Stories shape how we feel, what we believe and what we buy.
AI labs can't access licensed video
Tens of millions of hours of premium media sit behind rights that centralized labs can't clear, so their models never learn from it.
Captions aren't comprehension
Today's best vision-language models describe pixels. They miss narrative intelligence: motive, subtext, culture and emotion.
We decode the stories locked inside millions of hours of IP.
Scryer-1, a vision-language model built for narrative
A Narrative-Intelligence Mixture-of-Experts. Four specialist experts each read the scene through one lens, and a router decides how much each one counts for every frame. The result is a structured reading of the story, small enough to run on a viewer's own device.
Tap an expert to see what it contributes.
The Projectionist Network
Independent nodes, called Projectionists, run Scryer-1 on media they are already entitled to watch. Footage never leaves the device. Only the structured reading does, and overlapping nodes cross-check it. That gives compliant access to licensed IP, indexed in real time at a fraction of data-centre cost.
Watch
Contribute screen time you already have.
Decode
Scryer-1 runs on-device and outputs a narrative graph.
Verify
Other nodes check the same scene. Disagreement is flagged.
Earn
Verified readings earn Reel Points when the index is queried.
Illustrative. Rates assume one scene per ~15 s of footage, a 92% verification rate and launch multipliers for sport and creator content. Final reward rules are published at network launch.
Rails for AI-native media and ads that get it.
One API for companions, remixes, shoppable video and scene-level ad targeting. Explore the proposed API with sample responses. This playground runs locally; early access is coming soon.
Press "Run sample" to see the response.
The science of narrative intelligence.
Research preview
Scryer-1 evaluations, litepaper and this demo.
Projectionist beta
First nodes, verification and Reel Points testnet.
Scene API early access
Partner studios and ad platforms query the index.
Open network
Anyone can run a node. Companion SDK ships.
Fair questions.
Does my footage get uploaded?
No. Scryer-1 runs on your device. Only the structured reading of a scene (characters, beats, emotion scores, products) leaves it, never frames or audio.
How is this legal for copyrighted shows?
Projectionists only decode content they are already entitled to watch, and the network stores descriptions of scenes, not the scenes. Rights holders can also license their catalogues directly for deeper indexing and share in the revenue.
How do you stop people faking readings?
Every scene is decoded by several independent nodes. Readings that disagree with the consensus earn nothing and lower that node's trust score.
What do advertisers actually get?
Scene-level context instead of keywords: mood, brand safety, products on screen and the best moments for a break. Moment Ads in the demo above shows the output.
When can I run a node?
The Projectionist beta opens after the research preview. Follow us on X to hear first.
Become a Projectionist.
Nodes open soon. Follow us on X to be first in line when the network goes live.
Studios, platforms, advertisers and researchers: tell us what you're building.