search

rhasspy / piper

A fast, local neural text to speech system
https://rhasspy.github.io/piper-samples/
MIT License
6.68k stars 489 forks source link

Training a new model, hoping to receive assistance #652

Closed xiasi0 closed 2 days ago

xiasi0 commented 2 days ago

I am training a new Chinese model for Piper because the pronunciation tone of the Piper project's Chinese model is incorrect. I have prepared a larger dataset. But what puzzled me was that Piper's training documentation mentioned that the CSV file of the dataset only has one text column, and I used the dataset to contain the text corresponding to the audio, as well as the corresponding pinyin and tone. How should I handle this CSV file to meet Piper's requirements for the dataset I hope someone can give me a hint. Thanks The training guide mentions that CSV format should be:

id|text

My dataset is in CSV format:

 ID|text|prosody

If there is no 'prosody', Will it lead to non-standard pronunciation?

xiasi0 commented 2 days ago

Can someone share the available 'pip list'?