An AI model that restores speech recordings by using advanced transformer techniques to clean up noise, reverb, distortion and other defects — even in severely degraded audio.
VoiceRestore is a speech-restoration model based on flow-matching transformers trained in a self-supervised way. It aims to improve the quality of degraded speech recordings by addressing a variety of common distortions, such as background noise, reverberation (echo), compression artifacts, and low sampling rates or other fidelity losses.