Holographix: a percept-first codec and network substrate for Large Vision Models and Large Audio Models over UDP

Holographix is a percept-first codec and UDP transport substrate for RGB images and PCM audio on impaired links, including loss, burst loss, jitter, reordering, duplication and intermittent connectivity. Each signal is represented as a coarse approximation plus an int16 residual; the residual is spread across a population of holographic chunks using a golden-step modular interleaving so that any subset of chunks contributes to the whole signal rather than to a single local region. Decoding treats missing residual samples as zeros and produces an anytime estimate from whatever chunks are available, avoiding all-or-nothing frame failures. Beyond graceful degradation, Holographix also supports controlled refinement: independent partial encodings of the same scene can be fused (stacked) over time and across mesh paths to increase signal-to-noise ratio and effective definition, and the refined estimate can then be re-encoded and propagated as a new chunk population. On the network side, chunks are addressed by content identity, using a mapping from holo://object URIs to content identifiers, framed over UDP and replicated by a lightweight mesh. The evaluation protocol considers PSNR or SNR as a function of the number of received chunks k, the variance across different subsets with the same k, and refinement curves as a function of the number of integrated exposures.