Geometric Iterative Retrieval for Neural Audio Codec Resynthesis
2026-08-19 • Sound
SoundMachine Learning
AI summaryⓘ
The authors look at how to better restore high-quality audio from simplified digital tokens created by neural audio codecs. Existing methods either guess the tokens outright or try to predict a continuous sound, but the authors find these approaches limited. They propose a new method called geometric iterative retrieval that uses the structure of the codec itself to improve the audio restoration step-by-step in a continuous space. Their tests show this method works better than previous single-step methods for both speech and music.
Neural audio codecsResidual Vector Quantization (RVQ)Discrete token predictionContinuous regressionGeometric iterative retrievalContrastive retrievalCodec restorationSpeech synthesisMusic generation
Authors
Leo Schmidt-Traub, Frédéric Berdoz, Luca A. Lanzendörfer, Roger Wattenhofer
Abstract
Neural audio codecs based on Residual Vector Quantization (RVQ) have become the dominant discrete representation for token-based general audio generation, yet resynthesizing high-quality audio from coarse codec tokens remains an open problem and bounds the fidelity of every system that generates them. Prior work has framed resynthesis as a choice between discrete token prediction and continuous regression. We argue that this dichotomy is incomplete and introduce geometric iterative retrieval, a paradigm that uses the RVQ layer hierarchy itself as a natural iterative decomposition in continuous codebook space. Rather than classifying over discrete vocabularies or regressing to a single target vector, our method performs contrastive retrieval in the codebook's geometric space. We evaluate our method on codec restoration tasks across speech and music, and show improvements over both single-pass token prediction and one-step regression baselines.