everyvoice synthesize from-text --filelist ( expects a psv type file with a header line )

[marctessier]

When trying to synth while using a "--filelist" it expects a "psv" type file with a header line in the format like this file for example: preprocessed/validation_filelist.psv.

A few things:

1) The "--help" for everyvoice synthesize from-text needs a better help message explaining what it expects as input / make more clear... Below is what we currently see when we run "--help"

│ --filelist -f FILE Synthesize all audio in a given filelist. Use --text if you want to just synthesize one sample.

2) Un-able to use a "plain" filelist . Where only sentences | utterences , one per line without a header line in the file. We get this message below that might not be clear to the end user why the synth did not work.

/home/tes001/u/TxT2SPEECH/miniconda3_u20/envs/EveryVoice_vocoder-match/lib/python3.10/site-packages/pytorch_lightning/utilities/data.py:104: Total length ofDataLoaderacross ranks is zero. Please make sure this was your intention.

[roedoejet]

A plain filelist works for me, I think this rather has something to do with the number of utterances being less than the batch size

[roedoejet]

please re-open if you still have this issue, but I have tested and this seems to work for me