Issues with diacritics in French

[CorentinJ]

Hello, I found some issues with the French "à" when using the espeak backend. I'm running 2.2.1 and espeak-ng-1.49.2.

Command line:

echo "Je vais à" | phonemize -l=fr-fr -b=espeak
ʒə vɛ

The "à" is simply removed. The same goes for é, I haven't tried other characters.

Python:

>>> phonemize('Je vais à', language="fr-fr", backend="espeak")
'ʒə vɛz aaksɑ̃ɡʁav '

The behaviour isn't consistent with the above, but it is problematic in a different way: here the "à" is explicitly spelled out as "a accent grave"

[mmmaat]

Hi, unfortunately this is a espeak-ng related issue we cannot deal with in phonemizer... I suggest you to directly open an issue there. (I tried with espeak-1.48, espeak-ng-1.49 and espeak-ng-1.50, same result each time)

$ echo "je vais à" | phonemize -l fr-fr -b espeak
ʒə vɛz aaksɑ̃ɡʁav 
$ echo "je vais à pied" | phonemize -l fr-fr -b espeak
ʒə vɛz a pje
$ echo "ô" | phonemize -l fr-fr -b espeak
oaksɑ̃siʁkɔ̃flɛks
$ echo "ô César" | phonemize -l fr-fr -b espeak
oː sezaʁ

[CorentinJ]

Indeed I was able to reproduce with espeak alone and on the latest version. I filed an issue here: https://github.com/espeak-ng/espeak-ng/issues/854

[mmmaat]

Thank you!