Open eschmidbauer opened 7 months ago
Depending on defined cleaners in your conf.yaml phonemizer might bi inserting combining tilde to text converted from characters to phonemes.
Also, I am debating whether normalizing the mel-spec is even necessary for this project.
Can you advise how to use a different text phonemizer?
Different as in, what is your use case?
Depending on defined cleaners in your conf.yaml phonemizer might bi inserting combining tilde to text converted from characters to phonemes.
Im curious how to define a different cleaner in the config
This file has all the info about adding new cleaners https://github.com/p0p4k/pflowtts_pytorch/blob/master/pflow/text/cleaners.py You can get some inspiration from https://github.com/p0p4k/CoquiTTS/tree/dev/TTS/tts/utils/text to modify depending on your use case. Also, the tilde character thing, you can just continue and drop it, might not be a big issue.
Hi, thanks for sharing this project! I noticed a small issue when running:![image](https://github.com/p0p4k/pflowtts_pytorch/assets/7139998/2fac6c50-4614-4e51-943f-f997f3e059b8)
python generate_data_statistics.py -i conf.yaml
Im getting an undefined symbol exception error. When i added debug info, i found that the function_clean_text()
is insertingUnicode Decimal Code ̃
I checked my dataset, andUnicode Decimal Code ̃
is not in the dataset anywhere I added these lines to confirm the issue was coming fromclean_text = _clean_text(text, cleaner_names)
Any help would be appreciated, thanks!