Voice Swap was not trained on any data that wasn’t “ethically gained.”
Read the bottom of their FAQ that lists the exact databases in question.
The couple of datasets they used on top of all the data they directly pay artists to consensually provide have permissive licenses that only require attribution for use, and gathered their information directly from a group of willing, consenting participants.
They are quite literally the exception to the rule of companies claiming they’re ethical, then using non-ethically sourced data as a base for their models.
That AI was trained on absolute mountains of data that wasn’t ethically gained, though.
Just because an emerald ring is assembled by a local jeweler doesn’t mean the diamond didn’t come from slave labor in South Africa.
Voice Swap was not trained on any data that wasn’t “ethically gained.”
Read the bottom of their FAQ that lists the exact databases in question.
The couple of datasets they used on top of all the data they directly pay artists to consensually provide have permissive licenses that only require attribution for use, and gathered their information directly from a group of willing, consenting participants.
They are quite literally the exception to the rule of companies claiming they’re ethical, then using non-ethically sourced data as a base for their models.