[thorum]

You might check out this list from espnet. They list the different corpuses they use to train their models sorted by language and task (ASR, TTS etc):

https://github.com/espnet/espnet/blob/master/egs2/README.md