A few months ago the crew at Corridor Digital attempted an experiment with the AI art program Stable Diffusion along with the video editing program DaVinci Resolve and another AI program called Dreambooth. It was essentially AI rotoscoping as they took footage of themselves to use to get the program to produce the animation they wanted. It was not well received. In my own review of the result, Anime Rock, Paper, Scissors, that it works as a proof of concept but wasn’t ready for action. There were a lot of glitches, including the hands. Badly done hands is a bad thing when your story is about an extreme version of rock, paper, scissors (or “crossblades” in the video”). Many other people also complained about the problems but also how using artificial intelligence, which isn’t exactly how that works since so much as to be programmed in just to get that result, is a potential problem to animators who feel like if this gets good enough the more greedy clients would replace them. This is similar to the AI concerns of the current writer’s strike in 2023, being replaced by computers just as robots took over many physical labor jobs.
Well, when it comes to the technical errors, the Corridor Crew listened, and tried to find ways to refine the process. Tonight I’m finally going to go over the results now that Anime Rock, Paper, Scissors II has just been released on YouTube, having first been released on their website. Let’s start by seeing what they learned from the positive and negative comments of the first installment and how they sought to fix them, then the end result, and then how they approached the sequel.
First up, as the video was released on their website first, the Corridor Crew put out a video about what they learned from the first one and the comments by fans. Co-Founder Niko Pueringer goes over why they wanted to try this out, what they learned from the previous successes and failures, and how they used what they learn when approaching the next chapter.
The use of Vampire Hunter D: Bloodlust to the extent they used it was one of the biggest controversies. Animators put a lot of time and effort into this and using AI to animate in their place is seen as an insult, that they could be replaced. You can get some of these programs to replicate someone’s style, though not the nuances and emotion. You know how Data kept trying to paint in Star Trek: The Next Generation but he was accused even by himself of missing the subtleties and adding his own emotion (which as an android he lacks most of the time) and thus was basically just a copy with no heart? That’s kind of AI in real life. By aping the animation studio’s style and that of the character designer it’s seen as a cop-out if not outright plagiarism. Excep tthey never said “oh, the guys who worked on that worked on this” but it’s still not doing any work themselves. So this time paying an artist to create the look of the show is a good step but in the end it’s still replacing the animators, and I can imagine they still aren’t happy about it.
By coming up with their art style, and still doing some animation to clear up the mistakes it’s a step in the right direction. My personal opinion is that AI art generators are good for visualization of the end product when a reference otherwise isn’t available. You don’t suddenly turn into Greg Land for using reference. It’s how he uses reference to basically trace over and the poor way he sometimes do so (plus the use of pornographic images as a base or using personal pictures without permission) that is the problem. Now I’m not sure how AI uses the example art styles but it only works as a base, not an end product not only morally but with current technological limitations. I don’t mind using it as a reference but there is something weak about using it as the end product whether it’s animation or still images. The crew had to take what was rendered and work with it further plus they hired an artist to develop the style rather than tell the computer to imitate a style and try to replicate it by tracing Unreal Engine produced shots and tracing over it, though in that case the Unreal stuff was just used as a base and they added and removed things when necessary.
Credit to Josh Newland for creating a unique look for the second installment. Apparently he looked at numerous action anime shows and not just one, creating a character model that was unique by adding his own style to the traditional elements of Japanese artwork.
So now that it’s up on their YouTube channel let’s look at the end product. Does Anime Rock, Paper, Scissors II fix some of the animation flaws of the original?
The animation errors are not going away. The first one I saw was under two minutes in, where the King is leading his younger sons to their first hunt. Look at how his necklace changes and disappears rather quickly. I’ve seen my share of hand drawn animation errors and that’s still worse than some of them. It may be that the AI is still having trouble with the “in-betweens”, the images between primary shots. Some of the old behind-the-scenes videos on the official He-Man YouTube channel goes into this a bit more when talking about the animation hiring process.
The story is somehow sillier with the addition of the backstory. I know that’s the point because it’s an anime parody but it kind of went past my limits, or maybe I just hate Philip that much. Who knows? At the five minute mark I had to stop a bit and then force myself through the rest of the story. As a story critic more than an animation one that’s going to matter to me, but it may be personal preference more than anything else. The overly dramatic performances worked for a short skit but with the stakes raised and about twice the length the writing and acting flaws are more evident to me. If I wasn’t doing a review after discussing the first one I probably wouldn’t have bothered. You’re welcome to disagree of course but it was just a bit much for me, folks. Can they tone it down without going too far the other way and losing the concept? Man, I hope so! Apparently I’m going to have to go through this again.
With that let’s go back behind the scenes as they look over their own work. It took a lot more work than “pop it in the program and let it do everything”, but was enough bonus work done for your point of view?
I’m of curious what would happen if someone using traditional hand-drawn animation, whether on the computer or a bunch of paper since I’m sure cels are out, could do something better. There are still glitches, like the medallions. They’re still overdoing the exaggerations. And they clearly need a lot of artists anyway, so would doing it the old way, like they did with Lackadaisy, end up being easier in the long run? That’s a challenge I’d like to see someone take on. As it is there’s still a lot of work involved but could this lead to a new way to make animation, and would it be a benefit to the industry or are the animators right to be concerned? I’ve seen so many low-budget works that look like it it’s hard to tell. The stuff coming out of Kartoon Channel rarely looks any good. You’d also be surprised how often computer assists are used these days to limit mistakes and speed up production with both good and bad results because time is still a factor in doing a good job.
Is this going to replace traditional animation? I’m not convinced yet. It is another way to make art but there are still so many shortcomings to overcome. CG hasn’t replaced traditional animation and new ways to do 2D animation at different budget levels have their advantages and disadvantages depending on the process. Bluey doesn’t exactly look like The Smurfs but it works for the stories they tell. I’m in favor of new ways to tell stories and clearly there’s more to getting a good end product out than “stick it in the computer” but as always time will tell. I just hope they calm the acting and sillier parts down a bit (not too much as this is a parody) when we get to the third installment. Don’t be so interested in the art side that you forget the story and performance, guys.
BW Media Spotlight 