Reproduce “AudioCodex: Streaming Autoencoders for Live Audio” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Marco Pasini, Javier Nistal, Cyran Aouameur, Stefan Lattner, George Fazekas
AudioCodex turns stereo audio into continuous latents at roughly 10 Hz. It uses a causal encoder and one-step magnitude decoder, plus a separate causal phase estimator. On one NVIDIA A100, decoding one latent step took 41.38 ms, or 2.06× real time. At dimension 256, gFAD favored AudioCodex over CoDiCodec, but reconstruction FAD was worse in the reported comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.