Reproduce “AudioLCM: Text-to-Audio Generation with Latent Consistency Models” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Huadai Liu, Rongjie Huang, Yang Liu, Hengyuan Cao, Jialei Wang, Xize Cheng, Siqi Zheng, Zhou Zhao

AudioLCM accelerates text-to-audio generation by distilling a diffusion teacher into a consistency model that maps noisy audio latents toward clean audio in a few denoiser evaluations. On the AudioCaps test set, it reports FAD 1.67 and RTF 0.003 at 2 evaluations, versus 1.56 and 0.190 for its 100-evaluation teacher. Measurements used one GPU, and quality is not uniformly superior.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.