GaboxR67

/

MelBandRoformers

Like 45

Hello GaboxR67,I am from Russia, and I have a proposal and a question for you. In Russia and all CIS countries, there is a very popular music genre called Russian chanson. It is known and listened to not only in Russia and the CIS but also abroad. The most popular performer of this genre is the group Butyrka, whose songs are known to millions of people around the world.Here is my question and proposal: could you develop a model specifically optimized for this genre and this group? The goal is for the model to bleed as little of the important instrumental into the vocal track as possible, keeping the instrumental quality at its maximum—without artifacts, muddiness, frequency cutting, or dryness. I am not asking for flawless studio quality; it is completely understandable that a neural network cannot produce perfect results. However, it would be amazing if the model could be specifically tailored to this music genre and the songs of this group.Therefore, dear GaboxR67, on behalf of all fans of this genre and this group, I kindly ask you to create a model for high-quality instrumental isolation for everyone's favorite music style. If you agree, I am ready to provide you with tracks for training in lossless FLAC format, and I can also point out exactly where neural network models mistake certain frequencies for voice and accidentally bleed them into the vocal track, if necessary.