Training method of singing sound conversion model, singing sound conversion method, singing sound conversion system and medium

The invention discloses a singing sound conversion model training method, a singing sound conversion method, a singing sound conversion system and a medium, which can be widely applied to the technical field of singing processing. According to the invention, the acquired single-person reading recording data is preprocessed to obtain the single-person voice data set, then the singing sound conversion model is self-trained through the single-person voice data set, and the single-voice data in the single-person voice data set is used as the training data to train the singing sound conversion model in the self-training process. And after the synthesized singing data is output and obtained, adding the synthesized singing data meeting the preset requirement in the synthesized singing data obtained by the last step of training to the training data, and continuing to train the singing sound conversion model obtained by the last step of training, so that the singing sound conversion model obtained by the training is m