Reverb plays a critical role in music production, where it provides listeners with spatial realization, timbre, and texture of the music. Yet, it is challenging to reproduce the musical reverb of a reference music track even by skilled engineers. In response, we propose an end-to-end system capable of switching the musical reverb factor of two different mixed vocal tracks. This method enables us to apply the reverb of the reference track to the source track to which the effect is desired. Further, our model can perform de-reverberation when the reference track is used as a dry vocal source. The proposed model is trained in combination with an adversarial objective, which makes it possible to handle high-resolution audio samples. The perceptual evaluation confirmed that the proposed model can convert the reverb factor with the preferred rate of 64.8%. To the best of our knowledge, this is the first attempt to apply deep neural networks to converting music reverb of vocal tracks.

The quantitative evaluation includes two tasks;
  Reverb Conversion: interchanging reverb of two different inputs. We evaluate our metrics with a comparison between target reverberated and interchanged samples. A higher value represents a better result in all the metrics used.

  De-reverberation: eliminating reverb of the target input. Values of the x-axis below are the percentage of bus send ratio (γ) set for mixing source and reverb factor. The unit for STOI is percent(%), and SRMR and SI-SDR are in decibel (dB). A higher value represents a better result in all the metrics used.

The listening test was conducted with twenty participants. The participants were randomly given one of two different test sets with twenty-four questions each. For each question, three samples are presented - a reference sample, which is the output of the proposed model, with two different samples, which are an input of the model and ground truth of the reference sample (GT).