“…The denoising problem has been previously tackled with various techniques, such as Wiener filtering, wavelets [3], and spectral substraction [4,5], which only affected stationary noise. The removal of clicks and thumps had to be treated independently by firstly detecting them and then interpolating the missing samples [1,6,7]. With the rise of machine learning, deep neural networks have been successfully applied for different kinds of audio restoration goals, such as speech enhancement [8,9], bandwidth extension [10], audio inpainting [11], and low-bitrate audio restoration [12].…”