Currently we are using the user provided language and voice when storing an entry in the Utterance database. Since T257659: add language and voice to wikispeech-server output Speechoid however returns both the language and the voice used as part of its response.
The values in this response should be used in the Utterance database instead to allow Speechoid to e.g. alias or deprecate a voice or to use other fallback mechanisms.
Optionally we might want to log if the provided value and the response values don't match since this is likely an indication of the config needing to be updated.