Couple of cool links they provide for WIP:<p>> Additionally, we have begun experiments on two upcoming major features:<p>- neural packet loss concealment:
<a href="https://arxiv.org/pdf/2205.05785.pdf" rel="nofollow">https://arxiv.org/pdf/2205.05785.pdf</a><p>- deep redundancy:<p><a href="https://www.amazon.science/blog/neural-encoding-enables-more-efficient-recovery-of-lost-audio-packets" rel="nofollow">https://www.amazon.science/blog/neural-encoding-enables-more...</a><p>These are using generative AI to plug gaps when a real-time audio packet goes missing in a couple of neat ways.
The audio war sort of faded out of the collective conscious. I have no idea what happened since at least 2016. Did opus win? Why haven't I seen anything by xiphmont for what feels like a thousand years?