AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

High-Fidelity Audio Compression with Improved RVQGAN

Rithesh Kumar, Prem Seetharaman, Alejandro Luebs, Ishaan Kumar, Kundan Kumar

NeurIPS 2023 · Image generation › Autoregressive and latent generation · approach: Diffusion audio synthesis · 101 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace ReLU with Snake activations (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 6.92 -> 9.12Improved
remove the multi-band STFT discriminator (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsobjective metrics: 1.82 -> 1.80 STFT distanceNo clear effect
remove adversarial losses (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 9.12 -> 1.07Hurt
remove adversarial losses (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsbitrate efficiency: 99% -> 62%Hurt
replace multi-period discriminator with single-scale waveform discriminator (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 9.12 -> 8.51Hurt
replace multi-scale mel loss with single-scale high-hop mel reconstruction (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 9.12 -> 7.68Hurt
use a codebook latent dimension of 8 instead of 2 or 256 (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 2.22 -> 9.12; 5.09 -> 9.12Improved
replace projected lookup with EMA codebook learning (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSI-SDR: 9.12 -> 8.33Hurt
apply quantizer dropout rather than no dropout (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbps and lower bitratesreconstruction quality across bitrates: 10.14 -> 9.74 SI-SDRMixed
increase maximum bitrate from 8 kbps to 24 kbps (ablation)Improved RVQGAN · 44.1 kHz audioSI-SDR: 10.75 -> 13.83Improved
remove balanced data sampling (ablation)Improved RVQGAN · 44.1 kHz audio, 8 kbpsSTFT distance: 1.82 -> 1.94Hurt
use Improved RVQGAN rather than EnCodec at the tested 24 kbps configuration24 kHz audio compression · 24 kHz audio, 24 kbpsSI-SDR: 9.69 -> 16.40Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy