This is already looking so good, and the video of their workflow shows that it really isn't that much effort (in terms of resources) compared to traditional. Of course, there's still some janky effects here and there but it's looking really good.
I think in a few years pretty much anyone will be able to just record themselves on a green screen and make all kind of crazy things, it's actually coming up much sooner than I expected (i would have said maybe 10 years a year ago) if improvements keep coming. Honestly the decision of open-sourcing and making stable diffusion readily available to people has been complete bonanza for the space. Innovation happens so much faster when lots of people can participate.
A week or so ago I also saw an animation made by Netflix (that got quite a negative backlash for using AI) where they actually focused on generating the backgrounds. Of course they didn't really go in depth on how they did it but it just shows it's also something that's possible. Put 2 and 2 together and you get a really powerful workflow.
I'm pretty sure in the next few years we are going to start seeing incredible amateur work that's going to be done fast and by small teams. Honestly, I couldn't be more excited!
This got a ton of flak on Twitter as well[0]. Corridor Digital is known for showcasing the work (and passion) of digital and effects artists and animators in their react series, but here they are using the very technology that's going to put those people out of work in a couple of years.
They also got into NFTs with Beeple a couple of years ago. You want to like these guys because they seem like fellow nerds but then stuff like this makes you realize they're just another corporation trying to feed on the hype.
>stuff like this makes you realize they're just another corporation trying to feed on the hype.
I see Corridor as a group of enthusiastic and knowledgeable people who used cutting edge technology to make what they wanted and they are doing this since forever, the video I linked also made me come to this conclusion, the fact that they set out to turn such silly concept into an anime is more about the fun than the money, i would feel extremely proud of such an achievement and that's probably what they feel too.
>the fact that they set out to turn such silly concept into an anime is more about the fun than the money
If that were the case, they could have hired any number of animators to do it, the quality would have been better and they would have had more control over the look and style (it looks the way it does because they trained the AI on a specific anime.) The only reason to use AI is because it's about the money - to not hire artists.
I can all but guarantee this is a service Corridor or a subsidiary company will be offering soon. This isn't a few nerds hacking away on some crazy tech just to see if they can make it work, this is a pitch to animation studios that it's almost time to fire their staff.
And that's perfectly valid from a business point of view, but as far as I'm concerned Corridor loses all of their nerd cred for doing it.
>they could have hired any number of animators to do it
The reason why you will feel so proud is that you make it yourself, the image generator is just a tool as the many others they used during the process; money will not give you such fulfilment.
>but as far as I'm concerned Corridor loses all of their nerd cred for doing it.
But to be fair, history will probably side with you. AI's going to eat the world regardless of what anyone says. I wish there was an ethical way to use it, but there isn't.
They didn't even use controlnet in their workflow (because it wasn't out yet), so there is already a lot of new stuff to experiment with without waiting 6 months that would probably make the process easier and/or more consistent.
Theres a lot of basic stuff they arent even doing, like using previous frames and motion compensating them to merge into generation. Ive kinda got that working locally.
I think in a few years pretty much anyone will be able to just record themselves on a green screen and make all kind of crazy things, it's actually coming up much sooner than I expected (i would have said maybe 10 years a year ago) if improvements keep coming. Honestly the decision of open-sourcing and making stable diffusion readily available to people has been complete bonanza for the space. Innovation happens so much faster when lots of people can participate.
A week or so ago I also saw an animation made by Netflix (that got quite a negative backlash for using AI) where they actually focused on generating the backgrounds. Of course they didn't really go in depth on how they did it but it just shows it's also something that's possible. Put 2 and 2 together and you get a really powerful workflow.
I'm pretty sure in the next few years we are going to start seeing incredible amateur work that's going to be done fast and by small teams. Honestly, I couldn't be more excited!