Abstract

This disclosure describes transmitting video as an extremely compact semantic representation—

plain text scene and motion descriptions, stick-figure or pose/motion tracks, timing and cut

metadata, occasional reference images, or layered combinations—and reconstructing an

approximate video at the receiver with a local or shared generative model. The receiver

synthesises frames from that sparse signal together with a generative prior. The design accepts

lossy and synthesised visual detail in exchange for reduced network transfer. It records an

auditable provenance distinction between received source data or transmitted assertions carried

in the bitstream and generated fill synthesised at the decoder; optional fidelity-critical overlays for

text, logos, instructions, or audio; adaptive layer switching under bandwidth constraints; and a

system-boundary energy evaluation plan that compares network, sender-analysis, and receiver-

generation contributions without assuming proportional joule savings from bit reduction. Broad

generative and semantic video compression already exists in academic and patent literature. This

draft documents a bounded set of combinations and operational packaging as prior art disclosure.

It does not claim novelty clearance, report measured performance, or recommend the method for

factual, medical, safety, evidentiary, or security-critical footage.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.

Download

Share

COinS