Model training method, music generation method, device and electronic device

Abstract

The embodiment of the present application discloses a model training method, a music generation method, a device, and an electronic device. Embodiments of the method include: preprocessing the music fragment file, determining the note matrix, the note context matrix and the note position matrix corresponding to each moment in the music fragment; splicing the note matrix and the note position matrix corresponding to each moment of the music fragment, Input to the first cyclic neural network; The note context matrix and the note position matrix corresponding to each moment are spliced ​​and input to the second cyclic neural network; the output of the first cyclic neural network is spliced ​​with the output of the second cyclic neural network, and input to the third cyclic neural network; based on the note matrix and the prediction matrix sequentially output by the third cyclic neural network, each cyclic neural network is trained to generate a music generation model. This embodiment provides a model capable of music generation, and using the model to generate music reduces the cost of music creation and improves the efficiency of music creation.

Description

technical field [0001] The embodiment of the present application relates to the field of computer technology, and specifically relates to a model training method, a music generation method, an apparatus, and an electronic device. Background technique [0002] With the rapid development of the cultural industry, the digital music business has always maintained a large demand, and the number of new music needs to be increased every year. [0003] In the existing way, music creation is usually carried out by music artists. However, due to the small number of music artists, there are fewer works created each year. Therefore, using this method to create music cannot meet the demand of the market, and for entertainment companies, the cost of music creation is relatively high, and the creation efficiency is relatively low . Contents of the invention [0004] The embodiment of the present application proposes a model training method, a music generation method, a device, and an e...

AI Technical Summary

Problems solved by technology

[0004] The embodiment of the present application proposes a model training method, a music generation method, a device, and an electronic device to solve the problem of unsatisfactory demand, high cost of music creation, and low creation efficiency caused by music creation by music artists in the prior art. technical problem

Images