This technology is still at the "well that's neat" phase but I wonder what this will enable us to do once it gets better and cheaper. Imagine low-budget movie makers creating photo-realistic...
This technology is still at the "well that's neat" phase but I wonder what this will enable us to do once it gets better and cheaper. Imagine low-budget movie makers creating photo-realistic scenes without having to pay for sets, extras, etc. I'm sure people will find some new and nefarious uses for this technology, but I'm excited at least by what it will allow creative people with time but no money to do.
I'm really excited for what advancements in face generation, speech synthesis, and procedural narrative are going to mean for video games in the next decade or two. The big open-world titles...
I'm really excited for what advancements in face generation, speech synthesis, and procedural narrative are going to mean for video games in the next decade or two. The big open-world titles currently on the market have scripted main and side questlines, with endless random fetch quests or similar spawning in certain locations (usually trying and failing to blend seamlessly with the rest of the game).
I think it's inevitable that these primitive "radiant quests" are going to mature into systems strong enough to drive games entirely on their own merits. Writing teams, character designers, and voice actors will be obsolete, though subsets of all three skills will still be required to train the algorithms replacing that human work. We're only beginning to enter the uncanny valley of endless quest generation. At some point we'll come out on the other side, and entirely machine-generated adventures will be indiscernible from the human-crafted ones we enjoy today.
Speech synthesis has already gone past the "well that's neat" phase, in my opinion. Google's technology like tacotron2 and wavenet enable near realtime procedural speech generation. All that's...
Speech synthesis has already gone past the "well that's neat" phase, in my opinion. Google's technology like tacotron2 and wavenet enable near realtime procedural speech generation. All that's needed is a wider industry adoption and a slightly more polished product. We're almost at a point where computer generated voices could replace 90% of voice actor work, especially for things like audiobooks or game characters. Imagine having hundreds of voice actors available to read your audiobook, all procedurally generated.
It was an awful movie, but S1m0ne was based around the inevitable future of film where an actress is completely CGI. Its remarkable to see theses proofs in action.
It was an awful movie, but S1m0ne was based around the inevitable future of film where an actress is completely CGI.
You can get code for GANs from kaggle or GitHub and just insert your image. Or even better, find a library in python. Along with that, you need a powerful computer or a cloud computing service....
You can get code for GANs from kaggle or GitHub and just insert your image. Or even better, find a library in python. Along with that, you need a powerful computer or a cloud computing service. Cloud won't be much expensive though.
This technology is still at the "well that's neat" phase but I wonder what this will enable us to do once it gets better and cheaper. Imagine low-budget movie makers creating photo-realistic scenes without having to pay for sets, extras, etc. I'm sure people will find some new and nefarious uses for this technology, but I'm excited at least by what it will allow creative people with time but no money to do.
I'm really excited for what advancements in face generation, speech synthesis, and procedural narrative are going to mean for video games in the next decade or two. The big open-world titles currently on the market have scripted main and side questlines, with endless random fetch quests or similar spawning in certain locations (usually trying and failing to blend seamlessly with the rest of the game).
I think it's inevitable that these primitive "radiant quests" are going to mature into systems strong enough to drive games entirely on their own merits. Writing teams, character designers, and voice actors will be obsolete, though subsets of all three skills will still be required to train the algorithms replacing that human work. We're only beginning to enter the uncanny valley of endless quest generation. At some point we'll come out on the other side, and entirely machine-generated adventures will be indiscernible from the human-crafted ones we enjoy today.
Speech synthesis has already gone past the "well that's neat" phase, in my opinion. Google's technology like tacotron2 and wavenet enable near realtime procedural speech generation. All that's needed is a wider industry adoption and a slightly more polished product. We're almost at a point where computer generated voices could replace 90% of voice actor work, especially for things like audiobooks or game characters. Imagine having hundreds of voice actors available to read your audiobook, all procedurally generated.
But can it realistically animate them?
There's this: Nvidia AI Can Generate Virtual Worlds From Videos Also, deep fakes. We're getting there.
It was an awful movie, but S1m0ne was based around the inevitable future of film where an actress is completely CGI.
Its remarkable to see theses proofs in action.
Is there anywhere I can play with this with my own images and slider selections?
I'd love to have an app for that.
You can get code for GANs from kaggle or GitHub and just insert your image. Or even better, find a library in python. Along with that, you need a powerful computer or a cloud computing service. Cloud won't be much expensive though.
I did find the link to the paper.
https://arxiv.org/pdf/1812.04948.pdf
I'll see what I can get from there.