Comments on: The future is procedural <cite>We are machines</cite> We're complex systems formed by a combination of N simple organisms, where N is a really really big number. Anyhow, I'm nowhere near as smart as you guys (there's a reason I don't work in the development industry), but it seems to me that the problem with procedural animation is simply one of scale. You need either a massive number of eyeballs, or a massive set of movement data to be able to train your algorithms. Now if I wanted to solve the problem, I'd be thinking more along the lines of "how do I turn algorithm training into a game" so that I can get the training data from Facebook users than "how can I get a massive set of animation data". The <em><i>Enders Game</i></em> solution. We are machines

We’re complex systems formed by a combination of N simple organisms, where N is a really really big number.

Anyhow, I’m nowhere near as smart as you guys (there’s a reason I don’t work in the development industry), but it seems to me that the problem with procedural animation is simply one of scale. You need either a massive number of eyeballs, or a massive set of movement data to be able to train your algorithms.

Now if I wanted to solve the problem, I’d be thinking more along the lines of “how do I turn algorithm training into a game” so that I can get the training data from Facebook users than “how can I get a massive set of animation data”. The Enders Game solution.

]]>
By: Procedural Generation and User-Generated Content, Part I: Intro, Myths, and Facts » #AltDevBlogADay/2011/03/20/the-future-is-procedural/#comment-1892 Procedural Generation and User-Generated Content, Part I: Intro, Myths, and Facts » #AltDevBlogADay Wed, 23 Mar 2011 07:40:57 +0000 Ah yes, I take your point. My view is that we *can* achieve at least some of what you mention by computational means. Certainly as an aide to the creator, rather than a replacement. You're example of emotion driving motion is actually one aspect of my PhD thesis (which is on performance and emotional expression in digital actors), so my current academic work is experimenting with this field. It's also what my indie game is about and will be what I'm doing for Namaste. So I think there's definitely a lot of rewarding effort to be had in procedural animation towards exactly the same goals you describe. If you think its inappropriate for a computer to try and help with those things then thats fine, but I care that these things really arent done very well in games and its my intention to work on that problem. I think you highlight a number of things that is exactly what drives me to explore them. Emotional expression through movement, how we percieve movement and what we understand from it. I don't think its a pure mechanical thing, but certainly having the mechanics right cant hurt. I guess the proof of the pudding will be if I can make a scene that is entirely procedural and that affects you in the manner intended based on my own creative choices. Lets not relegate procedural animation to walk cycles just yet. Ah yes, I take your point. My view is that we *can* achieve at least some of what you mention by computational means. Certainly as an aide to the creator, rather than a replacement. You’re example of emotion driving motion is actually one aspect of my PhD thesis (which is on performance and emotional expression in digital actors), so my current academic work is experimenting with this field. It’s also what my indie game is about and will be what I’m doing for Namaste.

So I think there’s definitely a lot of rewarding effort to be had in procedural animation towards exactly the same goals you describe. If you think its inappropriate for a computer to try and help with those things then thats fine, but I care that these things really arent done very well in games and its my intention to work on that problem.

I think you highlight a number of things that is exactly what drives me to explore them. Emotional expression through movement, how we percieve movement and what we understand from it. I don’t think its a pure mechanical thing, but certainly having the mechanics right cant hurt. I guess the proof of the pudding will be if I can make a scene that is entirely procedural and that affects you in the manner intended based on my own creative choices.

Lets not relegate procedural animation to walk cycles just yet.

]]>
By: Mike Jungbluth/2011/03/20/the-future-is-procedural/#comment-1862 Mike Jungbluth Tue, 22 Mar 2011 06:39:22 +0000 I believe the notion of "soul" is not a reference to a higher being, it simply reflects that what is missing is an understanding of body-language. How can a mime or a dancer describe emotions? How can a martial artist show flair? How can you instinctively tell if a tiger is hunting? We, as humans, are very much tuned into observing and interpreting motions around us, and unless the meaning of motion is not taken into consideration, procedural animation will forever remain robotic and uncanny. I believe the notion of “soul” is not a reference to a higher being, it simply reflects that what is missing is an understanding of body-language. How can a mime or a dancer describe emotions? How can a martial artist show flair? How can you instinctively tell if a tiger is hunting? We, as humans, are very much tuned into observing and interpreting motions around us, and unless the meaning of motion is not taken into consideration, procedural animation will forever remain robotic and uncanny.

]]>
By: Rachel 'Groby' Blum/2011/03/20/the-future-is-procedural/#comment-1856 Rachel 'Groby' Blum Tue, 22 Mar 2011 01:24:15 +0000 Michael: I dont buy the whole "soul" bit. We are machines. If you want to suggest something higher than that, I dont mind, but I certainly dont agree. I think the rules of motion and even of stylised motion are there to be understood. Just as the rules of proportion are there to be understood. I haven't heard a compelling argument yet why it will be fundamentally impossible to procedurally generate aesthetically pleasing motion. Certainly dont buy the "soul" argument. Please feel free to enlighten me with a solid argument. Michael: I dont buy the whole “soul” bit. We are machines. If you want to suggest something higher than that, I dont mind, but I certainly dont agree.

I think the rules of motion and even of stylised motion are there to be understood. Just as the rules of proportion are there to be understood. I haven’t heard a compelling argument yet why it will be fundamentally impossible to procedurally generate aesthetically pleasing motion. Certainly dont buy the “soul” argument.

Please feel free to enlighten me with a solid argument.

]]>
By: Dana V baldwin/2011/03/20/the-future-is-procedural/#comment-1846 Dana V baldwin Mon, 21 Mar 2011 19:46:32 +0000 The first time I noticed these content style changes was on a much smaller scale than the changes you note on recent games. I'm thinking of the changes from say Baulder's Gate (sandbox) to KotOR (theme park). More voices and cutscenes and a more linear story track. Now I think a good term for this modern shit would be to call them roller coasters. You get to ride along just like Space Mountain! The first time I noticed these content style changes was on a much smaller scale than the changes you note on recent games. I’m thinking of the changes from say Baulder’s Gate (sandbox) to KotOR (theme park). More voices and cutscenes and a more linear story track. Now I think a good term for this modern shit would be to call them roller coasters. You get to ride along just like Space Mountain!

]]>
By: Michael Jungbluth/2011/03/20/the-future-is-procedural/#comment-1840 Michael Jungbluth Mon, 21 Mar 2011 17:10:59 +0000 The GA/NN approach is what naturalmotion does and it takes a LOT of hand tweaking to get right. Anyone who has actually done GA/NN has the same experience. I'm not really sure what the most solid approach is yet. But I'm reasonably convinced that the principles of motion are reducible enough that we can perform realtime procedural motion that looks good. The GA/NN approach is what naturalmotion does and it takes a LOT of hand tweaking to get right. Anyone who has actually done GA/NN has the same experience.

I’m not really sure what the most solid approach is yet. But I’m reasonably convinced that the principles of motion are reducible enough that we can perform realtime procedural motion that looks good.

]]>
By: Gavan Woolery/2011/03/20/the-future-is-procedural/#comment-1838 Gavan Woolery Mon, 21 Mar 2011 16:23:43 +0000 I think we all agree there. I guess the real question is how far we could push those tools. If you look at the example link to the Epic stuff, that was very much enabling designers to create nice buildings from relatively simple sets of designer-specified rules. The results were convincing and the time saved must be pretty big. But its still a pretty complex system. So the question is how to make the interface to define rules simpler? CityEngine does a similar thing, but has a "facade editor" interface that allows you to define how a facade splits up from an example photo. Thats one good example of making the interface simpler. Its very much an area that needs more work though. Especially in the animation side. I think we all agree there. I guess the real question is how far we could push those tools. If you look at the example link to the Epic stuff, that was very much enabling designers to create nice buildings from relatively simple sets of designer-specified rules. The results were convincing and the time saved must be pretty big.

But its still a pretty complex system. So the question is how to make the interface to define rules simpler? CityEngine does a similar thing, but has a “facade editor” interface that allows you to define how a facade splits up from an example photo. Thats one good example of making the interface simpler.

Its very much an area that needs more work though. Especially in the animation side.

]]>
By: Niklas Frykholm/2011/03/20/the-future-is-procedural/#comment-1831 Niklas Frykholm Mon, 21 Mar 2011 11:56:43 +0000 Niklas, thats a very confrontational way of looking at it. The point of procedural for me, is to empower the creator to be more creative. Of course it also means that the creator can be anyone, not just an elite "artist", but thats a by-product. If you're happy modelling every single grain of sand on a beach, then thats fine. But I suspect many artists would prefer to just generate the beach and get on with the interesting stuff. A case in point I think works well to show this: Get hold of Cryeditor and go to the terrain and paint some trees with the tree painter. The patterns of trees are procedural, so you get variation. Now go and do the same thing placing each blade of grass and each bush and tree by hand. Now tell me which approach is more useful. The fact that you have a default procedural approach with hand-edits to make things work where you really need them to was the point of my post. Note that the worms stuff had procedural defaults and then let you hand edit the results to taste is a very powerful one. I view the procedural generation as a helper that opens up the process to everyone. Not as a replacement for artists. Niklas, thats a very confrontational way of looking at it. The point of procedural for me, is to empower the creator to be more creative. Of course it also means that the creator can be anyone, not just an elite “artist”, but thats a by-product.

If you’re happy modelling every single grain of sand on a beach, then thats fine. But I suspect many artists would prefer to just generate the beach and get on with the interesting stuff.

A case in point I think works well to show this: Get hold of Cryeditor and go to the terrain and paint some trees with the tree painter. The patterns of trees are procedural, so you get variation. Now go and do the same thing placing each blade of grass and each bush and tree by hand. Now tell me which approach is more useful. The fact that you have a default procedural approach with hand-edits to make things work where you really need them to was the point of my post. Note that the worms stuff had procedural defaults and then let you hand edit the results to taste is a very powerful one.

I view the procedural generation as a helper that opens up the process to everyone. Not as a replacement for artists.

]]>
By: Phil Carlisle/2011/03/20/the-future-is-procedural/#comment-1827 Phil Carlisle Mon, 21 Mar 2011 09:27:07 +0000 @Niklas Frykholm - I agree with you to a certain extent. I come from a background in art and programming. I am not the best artist, by any means, but I think I understand enough to get by. In addition to traditional art methods, I have done a fair deal of work with Illustrator, Photoshop, 3DS Max, and Maya. Procedural generation is not necessarily about reducing artists and making programmers more dominant. It is actually about empowering everyone to make games better, cheaper, and faster. Artists are ALREADY using procedural generation, whenever they touch a computer. When you do a brush stroke in photoshop, that is procedural generation - it is taking a high-level command, and turning it into low level data. When you do an extrude or lathe operation in a 3D program, that is procedural generation - it is creating a more complex mesh for the artist based on a simpler 2D curve. This might sound like an oversimplification of what PG is, but the line has to blur somewhere. That said, let's say that procedural generation always produced crappy, "programmer-style" graphics. Is this a bad thing? I don't think so -- it would still be great for prototyping games or for indie devs who can't afford artists. PG is already capable of producing things that artists can't, just as compilers are capable of producing optimized machine code that a human never could. It is not like compilers are made to replace programmers, they are made to empower them. I think PG has always had the goal of empowering artists. There are some tasks which are simply not realistic (or necessary) for artists, like producing giant cities by hand. Artists With all that said, I still think you are very, very right. There has been very poor communication between programmer, artists, and management when it comes to creating effective PG, and this has kind of given people the wrong impression of what PG is capable of. I am actually addressing a lot of these issues in my article (coincidentially -- it has already been written), out on the 23rd. :) @Niklas Frykholm – I agree with you to a certain extent. I come from a background in art and programming. I am not the best artist, by any means, but I think I understand enough to get by. In addition to traditional art methods, I have done a fair deal of work with Illustrator, Photoshop, 3DS Max, and Maya.

Procedural generation is not necessarily about reducing artists and making programmers more dominant. It is actually about empowering everyone to make games better, cheaper, and faster. Artists are ALREADY using procedural generation, whenever they touch a computer. When you do a brush stroke in photoshop, that is procedural generation – it is taking a high-level command, and turning it into low level data. When you do an extrude or lathe operation in a 3D program, that is procedural generation – it is creating a more complex mesh for the artist based on a simpler 2D curve. This might sound like an oversimplification of what PG is, but the line has to blur somewhere.

That said, let’s say that procedural generation always produced crappy, “programmer-style” graphics. Is this a bad thing? I don’t think so — it would still be great for prototyping games or for indie devs who can’t afford artists.

PG is already capable of producing things that artists can’t, just as compilers are capable of producing optimized machine code that a human never could. It is not like compilers are made to replace programmers, they are made to empower them. I think PG has always had the goal of empowering artists. There are some tasks which are simply not realistic (or necessary) for artists, like producing giant cities by hand. Artists

With all that said, I still think you are very, very right. There has been very poor communication between programmer, artists, and management when it comes to creating effective PG, and this has kind of given people the wrong impression of what PG is capable of. I am actually addressing a lot of these issues in my article (coincidentially — it has already been written), out on the 23rd. :)

]]>
By: Niklas Frykholm/2011/03/20/the-future-is-procedural/#comment-1824 Niklas Frykholm Mon, 21 Mar 2011 06:12:26 +0000 Animation is one of the arts with the deepest uncanny valley. With good animation you can sell a blob. With even slightly off animation, a gorgeously modelled character looks like a zombie. The mechanical basics of animation are so convoluted - they have to do with our muscles, bones and brains - simulating from first principles is incredibly hard. I think any approach that starts from there is doomed. Far more likely to work is intelligent combining of hand-authored or mo-capped animations. Sorry, but I've seen a lot of Ken Perlin's stuff, and while I have a lot of respect for his brains, I think it's a dead end. Gleicher & Kovar's stuff is far more practical and relevant, also Rune Johansen's work in Unity. Animation is one of the arts with the deepest uncanny valley. With good animation you can sell a blob. With even slightly off animation, a gorgeously modelled character looks like a zombie. The mechanical basics of animation are so convoluted – they have to do with our muscles, bones and brains – simulating from first principles is incredibly hard. I think any approach that starts from there is doomed. Far more likely to work is intelligent combining of hand-authored or mo-capped animations. Sorry, but I’ve seen a lot of Ken Perlin’s stuff, and while I have a lot of respect for his brains, I think it’s a dead end. Gleicher & Kovar’s stuff is far more practical and relevant, also Rune Johansen’s work in Unity.

]]>
By: Gavan Woolery/2011/03/20/the-future-is-procedural/#comment-1819 Gavan Woolery Mon, 21 Mar 2011 00:08:26 +0000 I think there's some room in the space between real artist driven animation and procedurally driven. I think we need more work before we know what really works in this area. Certainly right now I dont think there's anything massively compelling. But that doesnt mean there never will be. I think the fundamentals of motion are well known enough that even stylised motions can be procedurally generated. Given enough effort and the will to do it. Plus of course access to an animators brain for a reasonable length of time. They do seem to object to being dissected though! I think there’s some room in the space between real artist driven animation and procedurally driven. I think we need more work before we know what really works in this area. Certainly right now I dont think there’s anything massively compelling. But that doesnt mean there never will be.

I think the fundamentals of motion are well known enough that even stylised motions can be procedurally generated. Given enough effort and the will to do it. Plus of course access to an animators brain for a reasonable length of time. They do seem to object to being dissected though!

]]>
By: Michael Jungbluth/2011/03/20/the-future-is-procedural/#comment-1812 Michael Jungbluth Sun, 20 Mar 2011 17:53:06 +0000 Great post, finally getting some attention on focusing on cutting down the costs of content creation. Procedural content is great for creating lots of variety, but I find very hard to match the aesthetic qualities good artists can provide. If combined with tools and workflows to generate content to make the content creators and artists work actually simpler and easier, then yes that is the future. But when procedural content is used to over-complicate the workflow, as I have directly seen on a big game in the past, this can lead to a big fail and actually make something that took a long time, take even longer if you want it to look a certain way that is wasn't initially designed for. Great post, finally getting some attention on focusing on cutting down the costs of content creation. Procedural content is great for creating lots of variety, but I find very hard to match the aesthetic qualities good artists can provide. If combined with tools and workflows to generate content to make the content creators and artists work actually simpler and easier, then yes that is the future.

But when procedural content is used to over-complicate the workflow, as I have directly seen on a big game in the past, this can lead to a big fail and actually make something that took a long time, take even longer if you want it to look a certain way that is wasn’t initially designed for.

]]>
By: snake5/2011/03/20/the-future-is-procedural/#comment-1800 snake5 Sun, 20 Mar 2011 14:03:55 +0000