![]() ![]() The Apple researchers set out to improve voice quality when data in the target language is limited, and to enable more efficient cross-lingual synthesis. The quality of synthetic speech has improved dramatically with the development of neural networks, but this progress has come with the cost of high compute and huge amounts of training data. An Apple research team has published a study showing that data from speakers of languages other than the target language can be used to improve the voice quality of text-to-speech (TTS) systems.
0 Comments
Leave a Reply. |