This repository contains the official Generator and Discriminator models of bin2bin, described in "A Time-Frequency Generative Adversarial based method for Audio Packet Loss Concealment"
Below are some listening samples of the proposed generative inpainting method. LOSSY versions are corrupted with 40% of lost packets.