Someone used neural networks to upscale a famous 1896 video to 4k quality

this thing is nigh onto miraculous ...I've already purchased it

Awesome! I've been thinking about getting the Topaz AI bundle lately as they would significantly decrease the time and effort to perform various photo processing tasks and you get results that you'd never get with a conventional Photoshop/Lightroom workflow.

I see "Brad Lake" is listed in the credits on that Zappa movie poster...is that you I presume? I'm curious as to your involvement with that project. Am a big Zappa fan and have been waiting for that doc to get released for some time now.
 
Awesome! I've been thinking about getting the Topaz AI bundle lately as they would significantly decrease the time and effort to perform various photo processing tasks and you get results that you'd never get with a conventional Photoshop/Lightroom workflow.

I see "Brad Lake" is listed in the credits on that Zappa movie poster...is that you I presume? I'm curious as to your involvement with that project. Am a big Zappa fan and have been waiting for that doc to get released for some time now.
Yup that’s me, and the movie will be out very soon, the world premiere will be at South by Southwest next month, with the Hollywood premire on march 21. I havent yet seen it , but have spent a little time with director Alex Winter and Ahmet Zappa and am confident and excited for it . As you may know , the movie as well as the media preservation project of Zappa’s vault (the two are intimately connected was initially funded by Kickstarter
And I felt strong enough about it and the legitimacy of Alex as a legit documentary director and equally legit Zappa freak, so I went all in. More to come when I know details
 
They're also using neural networks to remaster old videogames that used pre-rendered backgrounds where the original assets are unavailable (think PlayStation 1-era and Final Fantasy, Resident Evil).

I'd like to see the tech applied to old SD TV shows too. Neat stuff.
 
Get the public to believe a politician or others did something nefarious that never actually occurred.
Do you watch CNN? They do it all the time. :) In fact all news is like that. I've been in the news a few times (including somewhat respectable publications like NYTimes), and every single time what I (and people I worked with) actually said and what was printed in the article were two completely different things. Nobody reads retractions though so we didn't ask for them - the bullshit they decorated the article with was actually favorable to us, just not in any way realistic or true. And that coverage was not political in nature, it was tech. I can imagine political coverage has nothing whatsoever to do with what actually happened, most of the time. That's when I realized that the news people think is "real" is mostly actually total horseshit written by people whose motivation nowadays is to drive clicks and eyeballs rather than report what happened.
 
They fixed the video today. The original was hilariously preposterous. With machine learning though, it's impossible to say if it's impressive or not from only a single example like this. The proof is in using the net on cases which are dissimilar from the learning set. It's easy to fool a neural net to do what you want by feeding it contrived input. In other words, if someone was expecting this to be like Captain Picard saying "Computer...enhance!", sorry but this falls far short :).
 
Back
Top Bottom