Hacker News Reader

Comment by kundan2510 | Hacker News Reader

Don't miss the overdub demo: https://www.descript.com/overdub and the demo video: https://www.descript.com/

Most impressive demo and marketing video of a product I have seen in a while.

Many of the best products seem to come when the founder has enough capital to self fund a solid base product versus needing “metrics” that at such an early stage often track nothing imo.

Whoa, that overdub demo is pretty crazy. The inflection in puts into the typed in content is pretty spot-on.

Edit: You can get it to go old school tts synthesis if you feed it kanji letters.

People will be able to create some serious deep fakes of audio recordings with Overdub. Where it will get really wild is when you can change the voice of the original recording into your "voice of preference" for listening purposes. Ie- instead of Ira Glass narrating This American Life, think what it will be like when you can listen to TAL with Morgan Freeman narrating it!

As stated in the article, Overdub forces you to read random sentences so you can’t just feed in hours of audio. That being said, this technology is quickly becoming commoditized and I suspect others won’t be as scrupulous.

Their video pokes fun at that: https://youtu.be/Maz0TVyTwiw?t=141

Adobe had demoed similar stuff back in 2016. [0] Not sure what happened to it after that.

[0] https://arstechnica.com/information-technology/2016/11/adobe...

How far can you go with overdub ? Can you just record someone for some time, and then just write a podcast (skipping the actual recording)?

That is a really great demo. I had way too much fun replacing the text with death ray in almost every example.

try: fff f f f fffffffffff ff f

Reply on news.ycombinator.com