Fig. 4From: Text-to-speech system for low-resource language using cross-lingual transfer learning and data augmentationTraining flow diagrams for our multi-speaker TTS models. Transfer learning from the source languages to the target language is used, where a are models using the different amounts of the English dataset and the Mongolian dataset, b is a model using the Japanese and Mongolian datasets, and c is a model using the entire datasets of both high-resource languages and the Mongolian datasetBack to article page