HUI-Audio-Corpus-German: A High Quality TTS Dataset

The increasing availability of audio data on the internet leads to a multitude datasets for development and training text speech applications, based deep neural networks. Highly differing quality voice, low sampling rates, lack normalization disadvantageous alignment samples corresponding transcript sentences still limit performance networks trained this task. Additionally, resources in languages like German are very limited. We introduce “HUI-Audio-Corpus-German”, large, open-source dataset TTS engines, created with processing pipeline, which produces high transcription alignments decreases manual effort needed creation.