Phonetic Enhanced Language Modeling for Text-to-Speech Synthesis
This work addresses robustness problems in text-to-speech synthesis for applications requiring high-quality speech generation, representing an incremental improvement over existing methods.
The paper tackles robustness issues in language model-based text-to-speech synthesis by proposing a phonetic enhanced language modeling method, which reduces error propagation and improves performance as validated by objective and subjective evaluations.
Recent language model-based text-to-speech (TTS) frameworks demonstrate scalability and in-context learning capabilities. However, they suffer from robustness issues due to the accumulation of errors in speech unit predictions during autoregressive language modeling. In this paper, we propose a phonetic enhanced language modeling method to improve the performance of TTS models. We leverage self-supervised representations that are phonetically rich as the training target for the autoregressive language model. Subsequently, a non-autoregressive model is employed to predict discrete acoustic codecs that contain fine-grained acoustic details. The TTS model focuses solely on linguistic modeling during autoregressive training, thereby reducing the error propagation that occurs in non-autoregressive training. Both objective and subjective evaluations validate the effectiveness of our proposed method.