High-Fidelity Generative Audio Compression at 0.275kbps

TL;DR

This paper introduces Generative Audio Compression (GAC), a novel approach that leverages semantic understanding and generative models to achieve high-fidelity audio reconstruction at ultra-low bitrates, surpassing existing codecs.

Contribution

The paper presents GAC, a task-oriented audio compression paradigm that uses powerful generative priors to drastically reduce bitrate while maintaining perceptual and semantic quality.

Findings

Achieves 0.275kbps bitrate for high-fidelity 32kHz audio

Outperforms state-of-the-art neural codecs in perceptual quality

Maintains intelligibility at 0.175kbps, 3000x compression ratio

Abstract

High-fidelity general audio compression at ultra-low bitrates is crucial for applications ranging from low-bandwidth communication to generative audio-language modeling. Traditional audio compression methods and contemporary neural codecs are fundamentally designed for waveform reconstruction. As a result, when operating at ultra-low bitrates, these methods degrade rapidly and often fail to preserve essential information, leading to severe acoustic artifacts and pronounced semantic distortion. To overcome these limitations, we introduce Generative Audio Compression (GAC), a novel paradigm shift from signal fidelity to task-oriented effectiveness. Implemented within the AI Flow framework, GAC is theoretically grounded in the Law of Information Capacity. These foundations posit that abundant computational power can be leveraged at the receiver to offset extreme communication bottlenecks--exemplifying the More Computation, Less Bandwidth philosophy. By integrating semantic understanding at the transmitter with scalable generative synthesis at the receiver, GAC offloads the information burden to powerful model priors. Our 1.8B-parameter model achieves high-fidelity reconstruction of 32kHz general audio at an unprecedented bitrate of 0.275kbps. Even at 0.175kbps, it still preserves a strong intelligible audio transmission capability, which represents an about 3000x compression ratio, significantly outperforming current state-of-the-art neural codecs in maintaining both perceptual quality and semantic consistency.