FAMIUM MoQ
Research implementation of draft-ietf-moq-transport-16 from Fraunhofer FOKUS, pairing a WebCodecs-based client with a moq-rs server. Notable for running MoQ natively end to end — capturing Unity and Unreal scenes, encoding through FFmpeg, and demonstrating sub-40 ms glass-to-glass latency.
Category
Libraries & SDKs
License
Open Source
Status
Experimental
Languages
Rust, Javascript
MoQT Draft Support
FAMIUM MoQ is Fraunhofer FOKUS’s research implementation of Media over QUIC — built for prototyping and evaluating object-based media delivery rather than for production deployment. It implements draft-ietf-moq-transport-16, pairing a WebCodecs-based browser client with a moq-rs server over WebTransport and HTTP/3.
Why this one matters out of proportion to its size
Almost every commercial MoQ deployment bridges to conventional contribution formats and uses MoQ only for the final delivery hop, a pattern we set out in Nobody Is Rebuilding for MoQ. FAMIUM is one of the few places MoQ runs natively from capture to playback.
The demonstration takes Unity game scenes, encodes via FFmpeg, and streams over WebTransport — and reports end-to-end latency below 40 ms.
That figure deserves to sit next to every commercial claim we track. The published production numbers cluster between 250 ms and 700 ms; this is roughly an order of magnitude lower. It is not a like-for-like comparison — synthetic content, a research network, no CDN, no scale requirement, and a game engine that can hand over a frame the instant it renders rather than a camera and a segmenting encoder. But the gap is large enough to be informative on its own, and it points at where the time actually goes in production stacks: not the transport.
Read alongside our comparison of what MoQ latency claims measure.
Also notable
Fraunhofer has shown the transport integrated into Unreal Engine and Unity, which is a use case largely absent from the broadcast-oriented vendor field — interactive and synthetic media, where the publisher controls frame production directly and object granularity can genuinely be per-frame.
FAMIUM was demonstrated at NAB 2026 (booth W2343).