>>150998429 (OP)
I loved how janky it was. I've seen worse where they didn't do enough generations. Sometimes the AI doesn't give them correct voices or they would say the wrong lines. The clothes and tone kept changing too. If people think this is gonna get better, its not. Its rooted in AI's quadratic scaling self attention problem. And the only way to solve that is changing the laws of reality.
>>150998538 >If people think this is gonna get better, its not.
If people think this is going to get better, they're wrong.
Compare it to AI videos from two years ago: these videos were made by a real person on YouTube using AI with watermarks.
Studios will have better AI, but can Disney make a 100% AI movie right now? No, but I'm pretty sure they can make some scenes that save a lot of work, which means fewer animators are needed.
>>150998538 >If people think this is gonna get better, its not.
Compare it to AI videos from two years ago: these videos were made by a real person on YouTube using AI with watermarks.
Studios will have better AI, but can Disney make a 100% AI movie right now? No, but I'm pretty sure they can make some scenes that save a lot of work, which means fewer animators are needed.
>>150998675
You know what, fuck you. Four years ago it could barely make the TF2 guys say Big Chungus isn't funny, this is exponentially closer than before.
>>150998538 >If people think this is gonna get better, its not.
It's done nothing but get better and I fucking hate it for that exact reason. Unlike these spamming fucks with their inexplicable hatred for artists I feel like this could basically make the internet unusable in just a few years if this fucking shit keeps up. I fucking hope it DOES slow down.
>>150998637
No. It just can't. Look up quadratic scaling problem on self-attention. The best they can do is come up with better architecture so vector connections are selective instead of all encompassing but it doesn't solve the problem. It just diminishes it.
I built an AI project for the institution I work for. this was a huge problem our team came across. Doesn't even matter how much GPU you feed it. There's no overcoming that wall unless you change how math works.