Hacker News Reader

This is amazing, though I think the voice is one of the hardest problems for deepfakes. I find there is a lot of inflection and tonality missing from synthesized voices. Even with a 'old TV broadcast' sound filter to mask some of that stuff, I think people are quite good at recognizing when a voice isn't totally human :) But the visuals in this production are incredibly convincing to me.

View on HN

This film is from 2019, and the tech has advanced substantially since then.

E.g. have you heard the voices from https://play.ht/ ?

Those capture inflection and tonality quite well, and are able to reproduce near-natural sounding laughter, pauses, and other human qualities. I wouldn't say it's perfect just yet, but give it another ~5 years and it will get there.

Wow.

and they are multilingual as well.

For me it was the first time to (afaik) to hear tts of this Quality in my native language (which is a major one)

If you have seen the Apollo 11 movie that was released a few years ago, the footage is all very familiar (and real).

They spliced the content in a way that is misleading

Reply on news.ycombinator.com