High-Fidelity Generative Audio Compression at 0.275kbps

2026-01-31

AI Summary

This work proposes a novel generative audio compression (GAC) paradigm that shifts the objective from waveform fidelity to task-oriented effectiveness, addressing the severe distortion inherent in conventional and existing neural audio codecs at ultra-low bitrates. By synergizing semantic encoding at the transmitter with scalable generative synthesis at the receiver, GAC achieves high-fidelity audio reconstruction under extreme bandwidth constraints. The approach uniquely integrates generative modeling with information-theoretic principles, embodying the “more computation, less bandwidth” philosophy to transcend the limitations of traditional waveform reconstruction frameworks. Built upon the AI Flow architecture and powered by a 1.8-billion-parameter generative model, the system delivers high-fidelity 32 kHz audio reconstruction at just 0.275 kbps and maintains strong intelligibility even at 0.175 kbps—achieving a compression ratio of approximately 3,000× and significantly outperforming state-of-the-art neural codecs.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Large language models for search

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.

Problem

• audio compression
• ultra-low bitrate
• semantic distortion
• high-fidelity
• generative audio

Innovation

• Generative Audio Compression
• ultra-low bitrate
• semantic-aware compression
• AI Flow framework
• information capacity