End-to-end singing synthesis method, computer equipment and storage medium

The invention discloses an end-to-end song synthesis method, computer equipment and a storage medium. The synthesis method comprises the following steps: S1, obtaining a lyric text, a corresponding music score pitch and a music score duration from a given music score; extracting corresponding semantic information from the lyrics and taking the semantic information as a part of model input; s2, converting the lyric text into a corresponding phoneme sequence, and inputting the phoneme sequence, a music score pitch sequence and a music score duration sequence into a first input end of a pre-constructed singing synthesis model; s3, inputting the semantic information into a second input end of a pre-constructed singing synthesis model; and S4, synthesizing the singing corresponding to the real person through the singing synthesis model. According to the invention, the naturalness and expressive force of the synthesized singing can be improved, and the method can be applied to any singing synthesis scheme. 本发明公开了一种