I am trying to run the code from the following link. However, I don’t know which are the right commit versions for both TTS and WaveRNN. Can anyone help me?
Welcome the forum
It will depend on the model(s) you plan to run. The details of the available models are given in a table on the wiki here:
You mention WaveRNN, maybe just because it’s mentioned in that particular notebook, but if you specifically want to use the WaveRNN one for the vocoder then I think it’s any of the last three that should work (ie the row mentioning 260k and below) and you’ll want the commit correspondingly linked to in the third column (“Commit”) for the TTS repo. I think you’ll need to check the details links for the WaveRN commits - sorry I haven’t got the whole answer but hope that sets you on a path where you can find the rest yourself.
Thank you very much for your answer, it was of great help!
I am not really sure why, but I did not see the commit column in the table. Turns out I have been using the correct commit all along, but I still didn’t manage to run the code. It seems that the notebook is trying to import some function that is not available at the current branch.
I’m receiving the following error message:
Traceback (most recent call last): File "new_mozilla_synthesize.py", line 3, in <module> from TTS.utils.text.symbols import make_symbols, symbols, phonemes ImportError: cannot import name 'make_symbols'
Currently I am using the following branches and models:
TTS branch: 20a6ab3
I’m guessing maybe this script is meant to be used with later models, therefore later branches, but I’m not quite sure.
Thanks again for your time!
Are you sure you’ve got the matching commit for the notebook too? I probably ought to have clarified that earlier. It looks like the version of the notebook from 20a6ab3 doesn’t call make_symbols.