DO NOT report audio problems. Issues with the voice cannot be addressed and reporting it only serves to push the error percentage up, which means that the Welsh course won't be able to exit the beta stage. This error percentage needs to drop below 5%.
The Ivona voice 'Gareth' seems to have fewer glitches, although Gwyneth doe snot seem to have very many, to be fair. Perhaps worth considering a switch to 'Gareth' before going fully live?
Dechrau is both a noun and a verb-noun. It can mean 'to begin/start', 'starting/beginning', 'a start' or 'a beginning', for example. It is important to realise that in Welsh a verb-noun can act as all of: a verbal noun (beginning); the equivalent of an English infinitive (to begin); and a noun (a beginning). Also, some verb-nouns are spelt in the same way as their related nouns.
I'm puzzled by the request to not report audio errors. Some of them have been fixed since I started, so evidently it is possible
There is nothing that we can do as a team to fix audio errors. The audio is automatically generated by the Ivona TTS software, developed by a company now managed as part of the Amazon organisation. Remember, too, that some apparent errors may be due to people's own equipment and network connections.