Hacker News Reader

>A lot of fantastical claims made in this article with little citation.

Not everything comes with a citation. Things first happen, then get studied, even if all the signs are there.

(Also, 'little citation' my ass. The post has links to video animation AI models, studios that already do this thing, papers and relevant technologies, etc. And it's just a post, not an exhausting monography on the subject)

>I have friends who work in animation and illustration professionally. Based off their predictions, the truth is likely far more mundane.

"It's difficult to get a man to understand something when his salary depends on not understanding it" as the saying goes.

The change will not be driven by animators, but by people hiring animators that decide to cut costs by not hiring them.

Subpar results wont be a problem, if they're still good enough. There will be people helping and people touching up the end results, but half or less of what there are now...

I.e.: " These AI models will likely just become tools that automate or simplify existing animation workflows for larger projects. This would definitely result in less labor needed for any given project."

Yes, but much more dramatically so, to the point that there will big loss of jobs in the industry.

As software has become easier and faster to write, the demand for it has only increased, and salaries have gotten higher and higher.

I see no reason this couldn’t happen for animation. Studios demand more immersive, interactive experiences that require more and more detail. More animated content is generated. Every marketing team now needs 3D animators to make social media content, etc.

This seems like the logical progression and doesn’t necessarily result in a loss of jobs.

>I see no reason this couldn’t happen for animation.

Because software had room to grow into all industries - which huge ROI, applicable everywhere.

Whereas animation has a capped target: it's content, of a specific type even, not a way to help other industries generate profit, like software is.

In a world where everyone is basically always consuming content I don’t really see this as a hard cap. Even if you’re not mindlessly scrolling through social media or watching Netflix, you still might be watching tutorials for work, or talking on slack, where we already use gifs and emojis. It’s definitely possible that animation eats the world, it would just have to move further down the spectrum from content to regular communication.

Economic growth has plenty of upside. I’d not be surprised to see 5-10% yoy economic growth as we approach the AGI frontier.

I see the animation industry as being very similar to the VFX industry, and the famous case of Rhythm & Hue winning an Oscar award for what they did on the movie "Life of Pi" in the same year they went bankrupt and out of business.

The industry will still be there. But how much money you can make, and how many people you can afford to have on the payroll, that will diminish.

I think Corridor Crew is showing this effect with virtually every post to their YouTube channel.

> The change will not be driven by animators, but by people hiring animators that decide to cut costs by not hiring them.

I have no doubt that management hopes that they will be able to cut costs through AI! My doubts come from the fact that the people most certain that this possible future will come to pass seem to be armchair experts outside the industry where people in the industry seem mostly excited about new tools and bemused with the IP issues around AI.

It's not like the last 20 years have been free of disruptions from software tools? I have no more of a crystal ball than anyone else, but this is only one of many confident predictions about AI killing human experts that has not yet come to pass. It's certainly possible, but I wouldn't invest based on it.

I think it would be a big mistake to discount Jevons paradox here. It's very possible that the lower cost of images and animation would greatly increase their demand, and customers would still want a person to be part of this process.

...or a bigger pie as it becomes a smaller gamble to launch an animation project...

I think this is probably too limited a view. My guess is that animation becomes more like photo-editing. There are still experts who do it professionally and push the boundaries of the form, but it’s also a skill that’s expected for marketing jobs, graphic designers, web devs, etc.

In most cases "bigger pies" are just pipe dreams, and it's all about powerlaw distribution of profits to a few smaller players.

> These AI models will likely just become tools that automate or simplify existing animation workflows for larger projects.

The problem for artists is that the vast majority of the skills they learned as artists are useless when it comes to forming prompts. Assuming these models raise their quality and responsiveness a bit, and are surrounded with tools to fix expected problems (anatomy, symmetry, light, etc), I wouldn't expect a professional artist even with an aptitude for the tools to have much better results than an obsessed 14-year-old without a social life who's been playing around with them for a few months.

They're weird in that they in a way turn the process of creation over to the fans. It's like a declarative language for entertainment. While vision and creativity make artists stand out, the real skill for a artist is knowing how to use artists' tools. The real skill for the future painter probably is knowing how to use the AI tools.

People who typeset with metal had very few skills that transferred to InDesign.

The difference between the artist and the 14 year old is the former has artistic vision and style. If a 14 year old can master prompt engineering, so can the artist.

I simply don't think that's the difference. The difference is that the artist knows a bunch of craft. Between the people who know a bunch of craft, we're attracted to ones with particular artistic visions and styles, but don't know that average 14 year olds don't have the same range of artistic vision and style.

Doesn't this artistic vision and style boil down to taste, anyway? Claiming that the skill of the artist and the skill of the editor is identical is obviously wrong.

edit: in a way, making a work that targets a particular audience is an acknowledgement that the public has largely set the vision and style, and the artist is actually adept at working within it.

In my view the skill of the artist is to transfer an idea into the medium they work with, and if they work in a collaborative format that’s 50% of the job. It takes an enormous amount of skill to master the medium, and some mediums are easier than others. But the quality of the ideas they have matters a great deal. As does their ability to collaborate effectively.

I predict the end of commercial digital content. AI will eventually be generating what I want, not assisting create what Disney or Activision can sell me.

I am working on a tool to make that happen.

Religious traditions are not sacrosanct, and neither are contemporary civic life traditions. I don’t owe the past carrying on its modern patronage system of copyright maximalism and monopoly ownership of assets to extract rents from.

Doubtful every culture seems to impose demands on its people. Even the left is anti-free speech these days.

I think we'll see the opposite, a few elites using AI to generate the content they force everyone else to consume.

Perhaps there will be a black market for AI's, there already is one for illegally trained anime AI's.

A growing segment of our society is tired of serving cushy office workers.

I would not be so sure there isn’t a massive amount of potential energy among the public to put video game companies out of work and increase the pool of “essential workers”.

IMO software engineers should optimize for themselves by walking away from industry rather than give it ownership of their intellectual output to be wielded against them after code writing AI is trained against their work.

Y’all thought you were so smart but hubris often comes with the later realization the effort was used against them. If society can generate content without paying someone it will.

It's already happening. Auto-tweening has basically displaced hand animation for most TV cartoons.

Is that AI driven, or is it just "here are the control points on the key frames, make them go"?

It would be neat to be able to have fewer key frames and better in betweening so that more storytellers could have access without a team of CalArts grads. That doesn't seem unreasonable given what we've already seen, but I'm neither an animator nor an AI expert so I'm curious.

Both. It's currently mostly interpolation driven, but there are already added smarts and heuristics in the algorithms, and AI work is done to make it even better looking.

This has been true for decades and has nothing to do with “AI,” it’s just classic computer graphics.

It's algorithmic replacement of human artistic skill in the animation field. Nobody said it was AI, it's an indication of how flexible the industry can be when it comes to cost cutting.

100% animation of each frame died a long time ago, because people don’t notice much about individual frames.

But as shown in the video you still need to hand tweak a great deal even if you don’t mess with individual frames. Suppose someone is jumping through a window, you need the sound of breaking glass to sync with the glass being broken.

It’s hard to explain why some things the article suggests are really easy and other stuff is difficult, but context is still a hard problem for AI. https://xkcd.com/1425/

Really?

Most or all top anime content is still hand drawn, even the in between frames.

Try to name 3 top anime that was completely hand drawn and released in the last 5 years. It’s extremely rare for even key frames to be completely hand drawn, let alone every frame.

Shirobako, Violet Evergarden, Edge runners

Mind you there is CGI in these animes, but the frames themselves are still hand drawn. How else would you expect them to be done?

There's no magic here.

Cartoons can do it because they just use rotations, translations and skews, anime does not have that luxury. The pro is that, imo, when done right, anime is just beautiful to see, mesmerising even.

> Auto-tweening has basically displaced hand animation for most TV cartoons.

That's got nothing to do with AI, it's just economics, and it's been like that for years. Everybody doing animation on the cheap wants to avoid drawing every frame.

Prior to digital animation you used other techniques, like recycling frames or tracing over existing drawings (Disney used to do this all the time in their films).

What digital animation really allowed to happen was 2D puppeteering, where you could rig puppets and animate them in two dimensions without having to draw the key poses or the in-betweens. Now we're at the point where the standard way to do 2D animation is via puppets, rather than any sort of hand-drawn (be it key poses or full tweening)

I can't seem to substantiate this for anime. There was an app launched in 2018, and it's still priced at $399/seat/year, not acquired by neither Autodesk nor Adobe. I suppose the distinction between cartoons and anime is substantial?

1: https://cacani.sg/product/cacani-teams/?v=24d22e03afb2

https://en.m.wikipedia.org/wiki/Inbetweening

I feel like pessimistic models like these often miss induced demand. That is with better tools that makes the job easier you open the way to produce more. Stuff that otherwise wouldn’t get animated actually does, by virtue of the craft becoming easier.

I find it odd (but predictable) that people miss this. We’ve seen proliferation of other specialized craft in the past as the tools get better, particularly in programing. And it is not like non-linear systems are that rare among people that are looking into AI and making predictions around it.

Animation is an amazing craft, our culture would for sure benefit from having more of it, and as more and more stuff gets animated, more people are probably gonna get jobs doing it, despite the fact—or even because of the fact—that every individual can produce more of it then before.

Neural Radiance Fields (NeRF) alone by themselves will alter VFX fundamentally even without all the Stable Diffusion 2.0 (now with Depth!) etc generation.

NeRF lets you pick ANY smooth camera path in a 3D space built up with nothing more than a mobile phone. That's today.

These two phenomena will start to merge and the impact will be as transformative to the software side of media creation as the switch to digital from celluloid was for hardware. There was a time some still remember when not everyone had a camera in their pocket.

NeRF only works with static scenes, requires perfect camera pose, clean scene with clear edges and low exposure. For outdoors any tree with leaves is going to add a lot of noise and blur. For a natural scene indoors the camera exposure is going to be too high. While it is certainly a breakthrough, even “just” static scene reconstruction is still not solved, given the limitations above.

Note also, NeRF is not an “AI” model, it is more of fancy compact storage for objects, images or scenes.

Three years ago, our SOTA image generation models had very strong limitations (when compared to today's models), it can be difficult to see when and where the next breakthroughs will come - NeRFs are not there yet, but we may not be far away

NeRFs feel like where GAN's were about 3 years ago.

Maybe someday AI art will solve the problem of everything these models generate having greebles.