IMPACT: Iterative Mask-based Parallel Decoding for Text-to-Audio Generation with Diffusion Modeling

TL;DR

IMPACT introduces an innovative text-to-audio generation framework that combines iterative mask-based decoding with diffusion models, achieving high audio quality and fidelity with faster inference than existing diffusion-based methods.

Contribution

The paper presents IMPACT, a novel approach that integrates mask-based parallel decoding in a continuous latent space to improve speed and quality in text-to-audio synthesis.

Findings

Achieves state-of-the-art Fréchet Distance and Fréchet Audio Distance metrics.

Significantly reduces inference latency compared to prior diffusion models.

Maintains high audio fidelity while enabling faster decoding.

Abstract

Text-to-audio generation synthesizes realistic sounds or music given a natural language prompt. Diffusion-based frameworks, including the Tango and the AudioLDM series, represent the state-of-the-art in text-to-audio generation. Despite achieving high audio fidelity, they incur significant inference latency due to the slow diffusion sampling process. MAGNET, a mask-based model operating on discrete tokens, addresses slow inference through iterative mask-based parallel decoding. However, its audio quality still lags behind that of diffusion-based models. In this work, we introduce IMPACT, a text-to-audio generation framework that achieves high performance in audio quality and fidelity while ensuring fast inference. IMPACT utilizes iterative mask-based parallel decoding in a continuous latent space powered by diffusion modeling. This approach eliminates the fidelity constraints of discrete tokens while maintaining competitive inference speed. Results on AudioCaps demonstrate that IMPACT achieves state-of-the-art performance on key metrics including Fr\'echet Distance (FD) and Fr\'echet Audio Distance (FAD) while significantly reducing latency compared to prior models. The project website is available at https://audio-impact.github.io/.