Support NeoGAF

[efyu_lemonardo]

I'm usually a critic of AI and its overuse. However, if Turntable does in fact work as advertised, I think this is an exception, and a case of real benefit to the artists.

ProjectTurntable for Adobe Illustrator is capable of rotating vector images without warping the design elements.

This is the kind of tool that is instantly usable for game developers and graphic artists. It can be seamlessly and immediately incorporated into any 2D pipeline.

Adobe introduced Project Turntable along with several other tools in a recent presentation, but I think this one is the most impressive.

For more tools, check out Adobe MAX Sneaks 2024.

I haven't watched them all, so some may be duds. Here's another one that looks useful:
ProjectRemixALot can digitize sketches and intelligently change the layout and design to fit different shapes and proportions.

Share your thoughts, or post competing tools you've seen or used.

The Verge also has a piece on this technology, complete with gifs which I've quoted above, but I encourage you to just watch the Turntable demonstration. If it really works this well for many vector images, then this is a genuine game changer.

Edit: added poll

 

[IAmRei]

Competing tools : let's real human work by themselves.

AI can led to cheap graphic saturation and will disturb market. Even without AI there are millions of scam or awful games in the market. And imagine if the tools are there...

 

[Gonzito]

 

[efyu_lemonardo]

Competing tools : let's real human work by themselves.

AI can led to cheap graphic saturation and will disturb market. Even without AI there are millions of scam or awful games in the market. And imagine if the tools are there...

I don't think there's a single human graphic artist or animator who wouldn't jump at the opportunity to use a tool like Turntable. It doesn't replace the artist, it empowers them.

 

[StueyDuck]

Competing tools : let's real human work by themselves.

AI can led to cheap graphic saturation and will disturb market. Even without AI there are millions of scam or awful games in the market. And imagine if the tools are there...

how is it any different to your generic UE megascan game or your Unity asset store packs that games already use.

this could maybe at least let the unskilled put their imagination into things rather than just buying the same asset bundles off humble bundle

 

[hinch7]

Damn this is impressive. Can think of the implications for anime and animation. Awesome.

 

[ReyBrujo]

Competing tools : let's real human work by themselves.

The IT crowd (no pun intended) are the first ones to adopt AI in their development cycle. However that these tools exist does not mean they are going to be used blindly, the artists will still have to retouch certain frames or animations. Everyone complains about long development times but when technology able to shorten them arrives they reject it? Does not compute.

In the dragon example you can see there are times in the rotation where the shadows below its stomach aren't right, or the hands move slightly, all those should be fixed by the designer or the artist, however it's much faster to redraw a few sections of a few frames than redraw everything. This is pretty cool.

 

[POKEYCLYDE]

I like it. Doesnt get rid of the individuals creativity, but will cut down on the artists work time.

There's a similar tool that takes a 2D image and does it's best to make a 3D model of that image.

These 2 tools working in concert could save artists so much time while preserving creativity.

 

[Da1337Vinci]

Look impressive annd seems like to be a huge time saver.

 

[efyu_lemonardo]

Competing tools : let's real human work by themselves.

AI can led to cheap graphic saturation and will disturb market. Even without AI there are millions of scam or awful games in the market. And imagine if the tools are there...

I really do understand your hesitation. I'm usually a critic of AI and its overuse. However, if Turntable does in fact work as advertised, I think this is an exception, and a case of real benefit to the artists.

 

[navii]

The caption for the above reads “Terrible at graphic design? Just throw together a rough sketch of what you need and Project Remix A Lot can do the rest. GIF: Adobe”

Id say if you sketched the above you are better than 90% of graphic designers

 

[efyu_lemonardo]

The caption for the above reads “Terrible at graphic design? Just throw together a rough sketch of what you need and Project Remix A Lot can do the rest. GIF: Adobe”

Id say if you sketched the above you are better than 90% of graphic designers

Agreed, it's the less impressive of the two: useful for old fashioned designers who are better with a pencil than a mouse, but those are certainly a dying breed.
Other tools were also demonstrated, but I think Turntable is the most impressive.

 

[efyu_lemonardo]

I like it. Doesnt get rid of the individuals creativity, but will cut down on the artists work time.

There's a similar tool that takes a 2D image and does it's best to make a 3D model of that image.

These 2 tools working in concert could save artists so much time while preserving creativity.

Do you have a link to this other tool?

 

[POKEYCLYDE]

Do you have a link to this other tool?

The one I heard about is called Instant Mesh. It's on github.

GitHub - TencentARC/InstantMesh: InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models

InstantMesh: Efficient 3D Mesh Generation from a Single Image with Sparse-view Large Reconstruction Models - TencentARC/InstantMesh

github.com

 

[Holammer]

Content aware fill and now this. Adobe is killing it with AI.
Eventually I expect to be able to feed an AI a model sheet with your own drawings or AI generated ones and it'll create a 3d model rigged and ready for games, web comics or whatever you can imagine.

I'm seriously expecting a second golden age for web comics once this happens.

 

[Hrk69]

Say hi to your mom from me

The future is here

 

[NeoIkaruGAF]

 

[efyu_lemonardo]

I'm glad there isn't a consensus in the poll: let's hear some dissenting opinions.

Why do you believe this is a gimmick?

 

[Ponderling]

Aww a convenient little a.i supplement that saves the humans so much time, how adorable sure I'll buy it, can't wait for the next update:

 

[onQ123]

The real win will be when this can be done on the edge device/console side & games could be much smaller because the assets are 2D + 3D Splats that becomes 3D as the game loads .

 

[The Cockatrice]

When will people stop calling features/tools "AI" when it's not actually AI, at all and when will gaf stop believing it's "AI" lmao? Guess never. Idiocracy is no longer fiction. It's reality.

 

[efyu_lemonardo]

The real win will be when this can be done on the edge device/console side & games could be much smaller because the assets are 2D + 3D Splats that becomes 3D as the game loads .

That's a great idea which I hadn't thought of!
I could see Nintendo incorporating dedicated hardware into a device of theirs that does this.

I get the feeling from games like Super Mario 3D World that they already use some kind of acceleration for transforming basic shapes like circles and ellipses for perspective correct shadows.

Perhaps Switch 2 might even implement something like this in a future SDK.

 

[efyu_lemonardo]

When will people stop calling features/tools "AI" when it's not actually AI, at all and when will gaf stop believing it's "AI" lmao? Guess never. Idiocracy is no longer fiction. It's reality.

What would you call this, then?
My understanding is that there's an underlying neural network that was trained on a huge number of examples in order to be able to produce these results.

 

[ChuckeRearmed]

Good. More people will able to produce the content they want.

 

[SirTerry-T]

I don't think there's a single human graphic artist or animator who wouldn't jump at the opportunity to use a tool like Turntable. It doesn't replace the artist, it empowers them.

One right here...Adobe can get fucked.

Any artist worth their salt would and should be able to draw the same object in any orientation.
This is basically a tool for lazy fuckers.

 

[The Cockatrice]

What would you call this, then?

Anything but AI.

 

[efyu_lemonardo]

One right here...Adobe can get fucked.

Any artist worth their salt would and should be able to draw the same object in any orientation.
This is basically a tool for lazy fuckers.

If it was free and open source, would you still not use it in the interest of saving time?

 

[Davevil]

When will people stop calling features/tools "AI" when it's not actually AI, at all and when will gaf stop believing it's "AI" lmao? Guess never. Idiocracy is no longer fiction. It's reality.

 

[efyu_lemonardo]

Anything but AI.

Why not put some effort into your answer? How do you think this was programmed if not with a large neural network?

 

[Romulus]

I wonder how far away we are from someone with zero experience telling AI what to create and making a working game?

 

[The Cockatrice]

Why not put some effort into your answer? How do you think this was programmed if not with a large neural network?

The term has been bastardized that even a google search is now AI. "Neural network" is now a fancy word for good code.

 

[efyu_lemonardo]

"Neural network" is now a fancy word for good code.

No, it isn't. The term is reserved precisely for cases where it is out of reach - even for experienced software engineers - to construct an algorithm from the bottom up which could successfully tackle a problem with so many different variables.

I think it's pretty clear that object recognition and style recognition are integral parts of this tool's pipeline, and (to my knowledge, correct me if I'm wrong) the best results in those areas are achieved today by using deep learning methods.

Perspective transformation is also a field with years of research put into it, from applications such as robotic vision, motion tracking etc. And once again (correct me if I'm mistaken) deep learning has played a significant part in developing this technology.

So I repeat the question: what makes you so sure this is the result of an algorithm written from the ground up? And how would you even approach the problem of building such an algorithm without heavy use of deep learning?

 

[SirTerry-T]

If it was free and open source, would you still not use it in the interest of saving time?

Nope. I'm an artist I like to do it all myself...time constraints and all.
Adobe seem obsessed with pushing AI in all of their creative software suite which is the biggest irony of all this.

 

[efyu_lemonardo]

Nope. I'm an artist I like to do it all myself...time constraints and all.
Adobe seem obsessed with pushing AI in all of their creative software suite which is the biggest irony of all this.

I do appreciate your principles. But assuming one could produce identical results using this tool, I can't say I understand them.

I have friends who are animators and whenever we talk about such things they make it very clear that time constraints are holding back their ability to fulfill their visions.

 

[The Cockatrice]

No, it isn't. The term is reserved precisely for cases where it is out of reach - even for experienced software engineers - to construct an algorithm from the bottom up which could successfully tackle a problem with so many different variables.

I think it's pretty clear that object recognition and style recognition are integral parts of this tool's pipeline, and (to my knowledge, correct me if I'm wrong) the best results in those areas are achieved today by using deep learning methods.

Perspective transformation is also a field with years of research put into it, from applications such as robotic vision, motion tracking etc. And once again (correct me if I'm mistaken) deep learning has played a significant part in developing this technology.

So I repeat the question: what makes you so sure this is the result of an algorithm written from the ground up? And how would you even approach the problem of building such an algorithm without heavy use of deep learning?

I highly recommend playing both Talos Principle games as well as looking up the turing test. The term AI, as I said, has been bastardized to no end. AI is and will always be a machine capable of doing things independently of humans. "Deep learning, or "machine learning" isn't actually AI. A tool that does 10 thigs at the same time better isnt AI. A code that looks up the internet to give you answers isnt AI. A code that looks up a database for media and enhances them, or mixes them up to produce things isnt AI.

 

[Holammer]

Nope. I'm an artist I like to do it all myself...time constraints and all.
Adobe seem obsessed with pushing AI in all of their creative software suite which is the biggest irony of all this.

Panel from Ernie/Piranha Club by Bud Grace.

 

[efyu_lemonardo]

I highly recommend playing both Talos Principle games as well as looking up the turing test. The term AI, as I said, has been bastardized to no end. AI is and will always be a machine capable of doing things independently of humans. "Deep learning, or "machine learning" isn't actually AI. A tool that does 10 thigs at the same time better isnt AI. A code that looks up the internet to give you answers isnt AI. A code that looks up a database for media and enhances them, or mixes them up to produce things isnt AI.

Forgive my bluntness but this sounds more like a semantic disagreement at best, and at worst a misunderstanding of the meaning of the very terms you've used.

I don't need to play the Talos Principle when I've actually studied this field at an academic level, I'm sorry. Some of your definitions are plain wrong, and others are irrelevant to this example.

I suggest you start by looking up "machine learning" and understanding how it's defined. If you want to argue that there's a difference between human learning and machine learning (as it is defined), there is plenty of material on that debate as well. But the term "AI", as I've used it, and as is acceptable to use it these days, implicitly applies to machine learning.

 

[Bashtee]

The term has been bastardized that even a google search is now AI. "Neural network" is now a fancy word for good code.

It comes in waves and is nothing new. Topics that would have fallen under AI get decoupled from the term, because it isn't what people think AI is, then AI get's shunned. Rinse and repeat.

 

[Shodai]

It's very clear the future is not building and tweaking everything by hand.

It is also very clear that a new rendering paradigm is coming, where the traditional approach to 3D, mesh-based worlds will be overtaken by neural networks doing all rendering.

 

[yurinka]

I'm usually a critic of AI and its overuse. However, if Turntable does in fact work as advertised, I think this is an exception, and a case of real benefit to the artists.

ProjectTurntable for Adobe Illustrator is capable of rotating vector images without warping the design elements.

This is the kind of tool that is instantly usable for game developers and graphic artists. It can be seamlessly and immediately incorporated into any 2D pipeline.

Adobe introduced Project Turntable along with several other tools in a recent presentation, but I think this one is the most impressive.

For more tools, check out Adobe MAX Sneaks 2024.

I haven't watched them all, so some may be duds. Here's another one that looks useful:
ProjectRemixALot can digitize sketches and intelligently change the layout and design to fit different shapes and proportions.

Share your thoughts, or post competing tools you've seen or used.

The Verge also has a piece on this technology, complete with gifs which I've quoted above, but I encourage you to just watch the Turntable demonstration. If it really works this well for many vector images, then this is a genuine game changer.

Edit: added poll

I think this stuff obviously is going to be very valuable for artist, helping them to save a lot of time. Obviously won't do all the job for them, but will be a great help.

 

[efyu_lemonardo]

It's very clear the future is not building and tweaking everything by hand.

Definitely.

It is also very clear that a new rendering paradigm is coming, where the traditional approach to 3D, mesh-based worlds will be overtaken by neural networks doing all rendering.

This part I think is less clear at the moment. In the long term there will probably be some new paradigm based on a different way for creators to interface with their creation.

But a tool like Turntable is - to me - proof that we are not at that time yet.
What makes this tool so powerful is the fact that, given a vector image, it doesn't spit out a jpeg or PNG, but rather it returns vector images which the artist can then go on and continue editing, with complete control over the end result.

 

[Thief1987]

Why audience behave worse than children in kindergarten?

 

[Shodai]

Definitely.

This part I think is less clear at the moment. In the long term there will probably be some new paradigm based on a different way for creators to interface with their creation.

But a tool like Turntable is - to me - proof that we are not at that time yet.
What makes this tool so powerful is the fact that, given a vector image, it doesn't spit out a jpeg or PNG, but rather it returns vector images which the artist can then go on and continue editing, with complete control over the end result.

Whatever the final outcome, it feels like a necessity (if only for how long development timelines have become) and I suspect a lot of very smart people are working on approaches.

 

[efyu_lemonardo]

Why audience behave worse than children in kindergarten?

Probably because it's a promotional conference and many of them are just blind fans. But this thread is about the technology, assuming what was shown is not an exaggeration of its capabilities.
I do assume there is a limit to the complexity of the vector path you can input into the tool, before it starts making weird (inhuman) mistakes. But from the looks of it this can be useful if your style is not realistic.

 

[The Cockatrice]

Forgive my bluntness but this sounds more like a semantic disagreement at best, and at worst a misunderstanding of the meaning of the very terms you've used.

I don't need to play the Talos Principle when I've actually studied this field at an academic level, I'm sorry. Some of your definitions are plain wrong, and others are irrelevant to this example.

I suggest you start by looking up "machine learning" and understanding how it's defined. If you want to argue that there's a difference between human learning and machine learning (as it is defined), there is plenty of material on that debate as well. But the term "AI", as I've used it, and as is acceptable to use it these days, implicitly applies to machine learning.

You're completely wrong. Might as well call an oven an AI if you code its buttons to cook in 10 ways. If your "AI" can't pass the Turing test, it aint AI, simple as that. Lets agree to disagree then and move on.

 

[StueyDuck]

When will people stop calling features/tools "AI" when it's not actually AI, at all and when will gaf stop believing it's "AI" lmao? Guess never. Idiocracy is no longer fiction. It's reality.

I felt like this for a long time, often would explain it's just deep learning.

But with the common folk it's impossible to fight the vanacular, just give in and use the terminology.

It's like the grandparents who called all games and consoles "The Nintendo". It's easier to just say yes then explain.

 

[StereoVsn]

The tools certainly look great. I am though worried about impact to employment. On one hand it means indie game creators can have easier jobs to produce their vision.

On the other hand tools like these mean that large companies will reduce their workforce.

It’s happening with software devs, data analysts and so artists already. And these tools will drive further layoff mid to long term.

I know, I know progress and all that, but unlike previous technical waves AI means long term reduction in employment and our governments will not assist due to will or lack of financial capacity .

 

[efyu_lemonardo]

You're completely wrong. Might as well call an oven an AI if you code its buttons to cook in 10 ways. If your "AI" can't pass the Turing test, it aint AI, simple as that. Lets agree to disagree then and move on.

In the field of AI research there is a concept called an AGI (artificial general intelligence), which is not the same as an AI.
An AGI is an AI with human level capabilities - or better - across every domain of human achievement. This is obviously not something that currently exists. What does exist currently are AIs which are approaching human level performance in specific domains.

In other words, they are capable of generalization within some specific area and under certain limitations. This is what anyone using the term AI today means.

As those limitations become smaller, and as the areas in question are unified, we will approach an AGI. But that is not what anyone referring to existing tools is talking about today.

An oven that cooks in any number of pre-programmed ways is doing no generalization whatsoever, so it is not even a domain-specific AI.

This tool, on the other hand, is supposedly capable of transforming any vector drawing that fits within certain parameters of complexity and similarity to real or cartoon objects. That number is essentially infinite, and it includes an infinite number of cases that the software engineers at Adobe never explicitly taught the program to recognize.

More importantly, the way the engineers taught the program to transform vector images wasn't by telling it how to break that operation down into smaller and simpler previously defined computations, and then apply them in series or parallel according to pre-programmed logic.

What they did was write a program that could be taught to look at a huge number of examples, and then come up - on its own! - with smaller computational steps and with a way to apply them that generalizes beyond the examples it was shown.

This capability to come up with a generalization without having the rules explicitly given to it by a programmer is what defines AI and machine learning.

And what makes AI methods (such as neural networks, among others) so powerful is that they have been successfully applied in cases where even the programmers themselves don't know the explicit rules or the logic ordering them.

I hope that was helpful.

Regarding the Turing test - as originally envisioned by Alan Turing - it is nothing more than a philosophical argument that considered a particular application of AGI: its ability to imitate a human in a conversation.
LLMs are getting very good at this particular task, but they are clearly not an AGI because there are plenty of other things which they can't do at all such as drive a car, or draw a picture.

You could, if you wanted to, design a Turing-like test for a specific domain of human achievement, including the ability to take a simple 2D vector image and correctly extrapolate what it would look like from a new perspective, even when this perspective requires successfully guessing new information about the image.

If this Turntable tool is indeed as general as it is being presented in the video, then it would pass such a Turing test.

If you're interested in continuing this conversation I ask that you do so in a more constructive manner.

 

[lmimmfn]

It would be funny if this was faked and they were 3D models all along lol

 

[Fess]

The rotation is very impressive! I draw and rotating a character by hand is some of the most time-consuming procedures, and it’s rarely fun and it’s extremely difficult to keep the face looking the same.

But if it’s just Adobe and vector based art it’s not solving anything for me. I want AI rotation in Procreate.