Method for training voice conversion model, electronic device, and storage medium

A computer device, comprising a memory and one or more processors, the memory storing computer-readable instructions that, when being executed, causes the one or more processors to perform:

• acquiring a voice sample set, the voice sample set comprising a plurality of training voice samples that respectively correspond to a plurality of training object identifiers, one of the plurality of training voice samples corresponding to one of a plurality of pieces of voice content, at least one training voice sample in the voice sample set being a synthesized voice sample, the synthesized voice sample being obtained by performing voice sample synthesis on a preset voice content feature and a voice timbre feature corresponding to a training object identifier;
• extracting voice features of the training voice samples to obtain initial voice features that respectively correspond to the training voice samples;
• fusing initial voice features that belong to different training object identifiers and have corresponding voice content matching each other to obtain target voice features;
• obtaining a voice feature set based on the target voice features;
• and
• inputting a voice feature in the voice feature set into an initial voice conversion model to obtain predicted voice data, and adjusting a model parameter of the initial voice conversion model based on a difference between the predicted voice data and target voice data corresponding to a target object identifier until a first convergence condition is met, to obtain a target voice conversion model, input data of the initial voice conversion model and voice content corresponding to the target voice data matching each other, voice timbre corresponding to output data of the target voice conversion model and voice timbre corresponding to the target object identifier matching each other.

A non-transitory computer-readable storage medium, storing computer-readable instructions that, when being executed, causes the one or more processors to perform:

• acquiring a voice sample set, the voice sample set comprising a plurality of training voice samples that respectively correspond to a plurality of training object identifiers, one of the plurality of training voice samples corresponding to one of a plurality of pieces of voice content, at least one training voice sample in the voice sample set being a synthesized voice sample, the synthesized voice sample being obtained by performing voice sample synthesis on a preset voice content feature and a voice timbre feature corresponding to a training object identifier;
• extracting voice features of the training voice samples to obtain initial voice features that respectively correspond to the training voice samples;
• fusing initial voice features that belong to different training object identifiers and have corresponding voice content matching each other to obtain target voice features;
• obtaining a voice feature set based on the target voice features;
• and
• inputting a voice feature in the voice feature set into an initial voice conversion model to obtain predicted voice data, and adjusting a model parameter of the initial voice conversion model based on a difference between the predicted voice data and target voice data corresponding to a target object identifier until a first convergence condition is met, to obtain a target voice conversion model, input data of the initial voice conversion model and voice content corresponding to the target voice data matching each other, voice timbre corresponding to output data of the target voice conversion model and voice timbre corresponding to the target object identifier matching each other.