Conditional End-to-End Audio Transforms
This addresses audio transformation for applications like voice conversion and music synthesis, though it appears incremental as it builds on existing sequence-to-sequence architectures.
The paper tackles audio style transformation by developing an end-to-end sequence-to-sequence model that conditions on speaker identities or musical instruments to convert audio between multiple voices or instruments, achieving competitive performance on standard datasets.
We present an end-to-end method for transforming audio from one style to another. For the case of speech, by conditioning on speaker identities, we can train a single model to transform words spoken by multiple people into multiple target voices. For the case of music, we can specify musical instruments and achieve the same result. Architecturally, our method is a fully-differentiable sequence-to-sequence model based on convolutional and hierarchical recurrent neural networks. It is designed to capture long-term acoustic dependencies, requires minimal post-processing, and produces realistic audio transforms. Ablation studies confirm that our model can separate speaker and instrument properties from acoustic content at different receptive fields. Empirically, our method achieves competitive performance on community-standard datasets.