PRISM captures real-time state from any game. VTX stores it as an open, self-describing binary standard. Together: a universal data layer partners can build on without waiting for publisher integration.
PRISM is how we get the data out. VTX is how the rest of the world reads it. Licence PRISM. Build on top of VTX. Ship products other teams can't.
Every broadcast-grade data stream today depends on the game publisher either shipping an API or partnering for a server feed. PRISM doesn't. It runs locally on a spectating machine and reads state from live game memory, frame by frame. The title doesn't need to know we're there.
PRISM reads state directly from the running game process. Entities, transforms, weapons, abilities, scores, every frame. No SDKs, no waiting for the publisher to ship a data API.
The extraction layer is engine-agnostic. We've shipped PRISM across Unreal, Source, Hextech, RE Engine, and custom engines, for CS2, League, Rocket League, Dota 2, VALORANT, and StarCraft 2.
One spectator PC at a venue, or wherever the match is being played. No server access, no publisher integration, no new infra. Drops into existing production workflows unchanged.
Output is written live into the VTX format. Every downstream tool, whether it's a broadcast overlay, an AI pipeline, or a 3D Twin, reads the same stream. One capture, any number of applications.
VTX is an open, self-describing binary format for live game state. Designed to stream, to store, and to replay with identical fidelity. Built engine-independent in C++ so anyone can read it, from broadcast overlays to LLM training pipelines.
Every capture carries its own schema. Readers can load a VTX file from any title, any version, without a companion config. Forwards-compatible by design: new fields don't break old readers.
Same format, two modes. Stream live frames to a broadcast overlay with low latency, or persist full matches for replay and analytics. A tool written once reads either.
Jump to any frame of a stored capture instantly. Director scrubs it. Analysts query it. Models train on it. The format is indexed for it from frame 0.
Reference implementation is portable C++ with no engine dependencies. Integrate into Unreal, Unity, a web viewer, a Python analytics notebook, or any pipeline that can link a library.
VTX is open. Spec, reference reader, tooling. The format is meant to outlast any one product, any one vendor. The more places it lives, the more valuable every capture gets.
Entities, transforms, animations, events, team state, match state. Everything the game is doing, frame by frame. Not a summarisation, not a stats blob, the full state.
The defining VTX feature. Applications pick their layer: generic types for cross-game portability, contextual types for rich per-title depth. Same file, same reader, different amount of depth depending on what you're building.
The shared VTX vocabulary: entities, transforms, events. Build a tool against this layer and it reads any VTX capture identically, regardless of the source game. Perfect for cross-title dashboards, LLM pipelines, or any downstream that wants portability over fidelity.
Game- and genre-specific types layered on top: CS2 weapon data, MOBA ability casts, racing telemetry. Full per-title fidelity when you need it, without forcing every downstream tool to care about game-specific fields they'll never use.
PRISM + VTX is infrastructure. These are the product categories that sit on top of it today, with partners shipping against the same open format.
Every frame, every angle, every entity, live or on-demand. The missing data layer under every modern sports broadcast.
Structured per-frame state straight from live gameplay. Real ground truth for motion, decision, and simulation models, at gameplay fidelity rather than reconstructed video.
Any tool can read a VTX file. Analytics dashboards, second-screen companions, coaching tools, mods. Build on top.
Streaming-first binary. Low-latency data for real-time overlays, in-venue screens, and tactical feedback.
Full-state replays teams can scrub, annotate, and query. Not just heatmaps over a minimap, the actual positions and decisions, frame by frame.
Mobile companions that track a live match with structured data. Stat cards, per-player views, interactive timelines, all from a single VTX stream.
Combine VTX with the asset layer (UAF + Atlas) and you get a live Game Twin running in Unreal. The data side of the Twin stack is PRISM + VTX.
Ship it. The format is open, the reader is portable, the licensing is clear. If it can read a binary file, it can read VTX.
The VTX format spec, reference reader, and tooling are open source. PRISM is a licensable commercial product, but every capture it produces lives in an open container anyone can read.
Licence PRISM for live capture. Build against VTX for everything else. Let's talk about your use case.
Start a Conversation