Hacker News Reader

Comment by leobg | Hacker News Reader

How do they do it? Elevenlabs, AFAIK, requires you to verify your voice. and open source tools like Bark are nowhere near as realistic sounding.

View on HN

From the article: "Mr. Obama’s A.I. voice was created using tools from ElevenLabs, a company that burst onto the international stage late last year with its free-to-use A.I. text-to-speech tool capable of producing lifelike audio in seconds. The tool also allowed users to upload recordings of someone’s voice and produce a digital copy."

Well, then probably these audios must’ve been generated before Elevenlabs introduced their verification safeguards.

Seems like a good strategy. Do something irresponsible. Wait for the backlash. Then introduce safeguards. It’s free publicity, and gives your company that aura having something “so powerful we were forced to add safeguards“.

I use RVC, but the discord I used for support and trading models was taken down

They don’t require verification unless that’s changed very recently

How could any verification actually work without being faked itself?

It works up until someone has a tool that works well locally or a service which doesn't require verification, the existence of which would also answer this question.

They ask you to read a piece of text, displayed on the screen for a few seconds, and then they compare that voice to the voice of the audio data that you uploaded.

Step 1: find someone who is good at impressions.

Reply on news.ycombinator.com