News for hackers. Stuff that matters.

A comment on Voice clone of Anthony Bourdain prompts synthetic media ethics questions. Parent comment

Re:Voice clone of Anthony Bourdain prompts synthetic media e... by stainforth on Friday July 16, 2021 @08:43PM (#27862729)
How much time of audio or data do you need on average to train up something? I guess I'm wondering if I only had a few minutes of someone speaking, would that be enough?
[ Parent | on HN ]
Re:Voice clone of Anthony Bourdain prompts synthetic media e... by genewitch on Saturday July 17, 2021 @09:50AM (#27865756)
The implementation i am using works best with several <30 second clips. I've tested it by cutting up 20 minute interviews to only have the person i care about, and it seems to be about the same as a half dozen 15 second clips.
[ Parent | on HN ]