Singing synthesis method and device

Abstract

The embodiment of the invention provides a singing synthesis method and device. The singing synthesis method comprises the steps of obtaining a target song and to-be-simulated user voice; acquiring song features of the target song and acoustic features of the user voice; based on the song features, the acoustic features and a pre-trained synthesis model, determining a plurality of acoustic parameters for synthesizing the song; and synthesizing the singing sound of the user based on the plurality of acoustic parameters. Thus, the song features of the pre-processed target song and the acoustic features of the user voice can be obtained, the acoustic parameters are determined by using the synthesis model, and the singing sound is synthesized by using the acoustic parameters, so that automaticsinging sound synthesis can be realized through model learning in singing sound synthesis, the singing sound synthesis rate is improved, synthesis factors are rich, and the trueness of the synthesized singing sound is high.

Description

technical field [0001] The present application relates to the technical field of speech synthesis, in particular, to a singing voice synthesis method and device. Background technique [0002] With the continuous development of artificial intelligence technology, more and more attention has been paid to the ability of voice interaction, and the ability to listen, speak, and sing has become an essential skill. More and more application software and smart devices begin to use speech synthesis technology to simulate human speech content, such as using speech synthesis technology to simulate human speech content for the purpose of "learning the tongue", or using speech synthesis technology to simulate human singing, etc. Voice different from normal speaking scene. [0003] However, at present, in the scene of simulating user singing, the usual practice is to model the voice information of different users, use the established model to convert the timbre of the target user's voice...

AI Technical Summary

Problems solved by technology

[0003] However, at present, in the scene of simulating user singing, the usual practice is to model the voice information of different users, use the established model to convert the timbre of the target user's voice, and combine the user's timbre with the inherent timbre of the song Adjust the timbre of the song through manual and other empirical methods to achieve the singing effect of simulating the user's timbre. Not only is the processing speed slow, but also the inherent timbre is used for adjustment. The simulation results obtained are relatively dull and do not match the actual singing voice.

Images