Mastodon Feed
Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
emilyskidsister@hachyderm.io wrote:
There are literally 5 speakers labelled as transgender in the validated English Common Voice dataset. If you would describe your voice as trans, and you record a few sentences with a desktop mic, this might actually make speech recognition, transcription, and text-to-speech suck less for us.