@ianthetechie they haven’t ever released those models as open weights, just via their API

I expect they may announce improved API versions of their text to speech stuff today at their DevDay event