Hacker News Reader

Comment by fxtentacle | Hacker News Reader

Let's all take a moment to reflect on how much worse we (the software developers) have made this by building (essentially automated spam) tools like https://www.synthesia.io/ For example, their "Lays Messi" ad campaign had 650 million video variations available in 8 languages....

I predict that in a year, the majority of internet videos will not only be random stock 4K videos, but they will be overlay-ed with AI avatars gesturing to AI voice synthesis generated off GPT-3 text.

You'll be able to search for any topic and find pretty videos talking about it. But listening to it will rot your brain, because you're hearing the equivalent of lottery numbers.

View on HN

Yeah that's youtube shorts. One of the reason I use vanced even if I have youtube premium is that it's the only way to just turn that half baked feature off. I wouldn't mind youtube's tiktok clone as much if it at least didn't have such a weird UI. It's so badly integrated to the rest of the app that it completely messes with your navigation buttons, and in general looks like a separate app inside the app. Which means performance is very bad too.

Instagram pushes its own tiktok clone very hard too (Reels), but the experience is a lot smoother compared to YT Shorts.

> https://www.synthesia.io/

Dear lord, that page sucked the lifeforce out of me. My condolences to those poor souls who worked on that.

I've said this before, and I'll say it again: Space Ghost Coast to Coast is probably the best example of what this could end up as. Old animations recycled to match voice lines.

Given advanced enough AI, the process could definitely be automated to turn a written script into a short video.

In my opinion, that is way too creative.

I expect the future of automatically generated content to look like a attractive + friendly + diverse + optimistic face which is utterly emotionless and inhuman, yet perfectly symmetric.

Here's a good Chinese example of an AI news anchor: https://www.youtube.com/watch?v=5iZuffHPDAw

But there's still a world of difference between them, mainly w.r.t. head movements, body language, and eye expressions. For example, compare the AI avatar here: https://youtu.be/-2AIweN_Fmg?t=272

With the real person here: https://youtu.be/-2AIweN_Fmg?t=27

And Synthesia is so awful that Synthesia didn't use it to generate their promo video. That's about a damning as it gets!

Nothing employees love more than soulless, robotic training videos.

It seems as if they read the Dead Internet Theory and thought, "you know what, let's make that real"!

Reply on news.ycombinator.com