dragonelite
Member
Cool thanks for posting that, i learned a bit more about game asset pipeline.
I'm fully embracing this inevitable video game crash
$70-80 games next gen(still with on disc DLC)
I'm fully embracing this inevitable video game crash
I'm fully embracing this inevitable video game crash
I have a question on this then.
So these guys finding it doubling their dev costs must not make PC games where the textures and such are far higher in res and the polycount is increased correct?
It seems that if a dev is a PC dev they would already be using far higher res models and so forth. Well actually many areare, as they talk about it all the time when talking about prepping different versions and lowering the complexity and so forth for the current gen.
Maybe I am wrong. Just seems weird to read of the issue when it seems like many PC games are probably already using assets. Of course if it is a console only dev I can see that. Maybe that's why some are happy about the upgrade and some seem scared. Makes sense I guess.
I pretty much guarantee that Sony and MS would rather ride out the generation as long as possible, especially because HD game development is not as pricey as when the generation first started, but still a bit expensive. There are some developers already having problems with HD consoles. However, the market wants a push onto the next generation and that's when it gets expensive again.MS and Sony are taking a huge risk that may end up costing all of us a steady flow of quality console games.
If one is afraid of text, feel free to skip.
TL;DR. Artists are generally always wanting to do better than their last project. They always try to push push push. You give them extra resources they will take it, fill it and demand more. That's just how we are. Milk every drop.
Each company has different ways and pipelines for making assets. Figuring out a good pipeline to get a character from the concept phase to in game and working is crazy long.
I'll give you a basic pipeline from start to finish.
Concept Phase, pretty self explanatory. Batch of thumbs and designs till its whittled down to one core design.
Modeller gets that, talks to rigger/animator if there is any special needs this character might require.
Character modeller models a low poly base mesh. Passes this off to the rigger as a proxy. (if the rigger requests this for a special case)
Character Modeller then takes the base mesh to a high poly stage. This may require a re-topology stage during this (usually if they pre-planned their base mesh well or the mesh isn't very complex this isn't required). Where the they reconstruct a new base mesh with better poly flow to create a cleaner high poly.
Then character modeller gets the high checked off.
Character modeller then does a final re-topology of his high poly asset to a lower poly in game version. This is always LOD0 (highest polygon version) obviously. (this job can be palmed off to a specific artist who only handles retopologising, UV unwrapping and baking)
NB: This is the part where they will be doubling their polygon count for next gen machines.
Then comes the UV unwrapping of the low poly game mesh and packing of UV's
NB: This can take a little longer if you have more polygons to deal with. But its not so bad these days with the tools available.
Then you need your different bakes from your high poly source mesh. This can be quite a complex process if there are lots of fiddly overlapping objects on the character. Things need to be pulled apart and baked separately. Usually two maps will be baked. The normal Map and the Ambient Occlusion (AO)
You can get more like displacement, height and cavity maps baked if needed.
Then you go from there onto the texture phase. This can be handed over to a specific texture artist.
NB: This is where you may get higher texture budgets. Painting a higher res texture takes more time. Also if more maps are needed for shader tricks, then more time is spent making special case textures.
After that its setting up all the sharers for the character. Depending on your shader pipeline (something that is worked together between tech artist/FX artist and code)
Then once this has been done the rigger then gets the model.
NB: Shader complexity may increase for the artist to hook up. But this is all dependant on the specific shader pipeline of your engine.
The character modeller then makes the needed LOD's for that character.
The rigger or Tech character artist, sets up all the final rigs and controls. Usually sets up the collision and all the technical gizmos that are needed like Sockets and code controls.
NB: Next gen machines may allow more bones per chunk. So this may add more time in the rigging phase to set up.
Then this is passed over to the animator for animation. Blends, animations and all that.
Then the animations need to be hooked up to the game design and code. So we can say when SPACE BAR is pressed, character plays Jump animation. ect ect.
NB: With more bones per chunk, Animators may need to spend longer amounts of time animating (Yes there is mo-cap which has its own stupid crap to deal with. cleaning up keyframes.. bleh. Also it can't be used for lets say, a crazy squid monster of doom). Though I am not an animator, so I'm only talking from observation.)
Then the character is finally in the game.
Wow finally its in the game! Oh shit... what happens when, for example, you show this to the publisher and he decides you character needs to wear a cloak? Then you have to dismantle the pipeline back to the modelling phase to add the cloak. the rigger/character tech artist then needs to re bake weights and add the new bones and re rig the bastard. Then the animator needs to add all these cloak animations to his existing animations.
One fuckup or one change can bring the whole pipeline to its knees. The more complex it is the harder the pipeline crashes.
Do this times X number of characters in a game. It can add up.
Now environments that's a different story. You usually only need to do up to the collision part. (Unless you had to make a complex prop with animation of course) We need to make collision meshes (soooo boring) and try to find new ways of saving drawcalls while adding more to a scene. Environments of course need a variety of shaders for different materials and different shaders on the one object also means more drawcalls. For example a tree will need two shaders because wood behaves differently than leaves. Its a discipline quite different from characters but takes just as long because the basic process is the same with the baking and the source models.
NB: Artists are generally always wanting to do better than their last project. They always try to push push push. You give them extra resources they will take it, fill it and demand more. That's just how we are. Milk every drop. If a frame rate is shitty in a game its because the artists pushed too hard expecting code to keep up.
For reference I have been both a character and environment artist for a next gen single player game utilising unreal engine 3 and also a next gen MMO. The pipeline I described was what I used when making characters for the single player unreal3 game.
I pretty much guarantee that Sony and MS would rather ride out the generation as long as possible, especially because HD game development is not as pricey as when the generation first started, but still a bit expensive. There are some developers already having problems with HD consoles. However, the market wants a push onto the next generation and that's when it gets expensive again.
If one is afraid of text, feel free to skip.
TL;DR. Artists are generally always wanting to do better than their last project. They always try to push push push. You give them extra resources they will take it, fill it and demand more. That's just how we are. Milk every drop.
Each company has different ways and pipelines for making assets. Figuring out a good pipeline to get a character from the concept phase to in game and working is crazy long.
I'll give you a basic pipeline from start to finish.
Concept Phase, pretty self explanatory. Batch of thumbs and designs till its whittled down to one core design.
Modeller gets that, talks to rigger/animator if there is any special needs this character might require.
Character modeller models a low poly base mesh. Passes this off to the rigger as a proxy. (if the rigger requests this for a special case)
Character Modeller then takes the base mesh to a high poly stage. This may require a re-topology stage during this (usually if they pre-planned their base mesh well or the mesh isn't very complex this isn't required). Where the they reconstruct a new base mesh with better poly flow to create a cleaner high poly.
Then character modeller gets the high checked off.
Character modeller then does a final re-topology of his high poly asset to a lower poly in game version. This is always LOD0 (highest polygon version) obviously. (this job can be palmed off to a specific artist who only handles retopologising, UV unwrapping and baking)
NB: This is the part where they will be doubling their polygon count for next gen machines.
Then comes the UV unwrapping of the low poly game mesh and packing of UV's
NB: This can take a little longer if you have more polygons to deal with. But its not so bad these days with the tools available.
Then you need your different bakes from your high poly source mesh. This can be quite a complex process if there are lots of fiddly overlapping objects on the character. Things need to be pulled apart and baked separately. Usually two maps will be baked. The normal Map and the Ambient Occlusion (AO)
You can get more like displacement, height and cavity maps baked if needed.
Then you go from there onto the texture phase. This can be handed over to a specific texture artist.
NB: This is where you may get higher texture budgets. Painting a higher res texture takes more time. Also if more maps are needed for shader tricks, then more time is spent making special case textures.
After that its setting up all the sharers for the character. Depending on your shader pipeline (something that is worked together between tech artist/FX artist and code)
Then once this has been done the rigger then gets the model.
NB: Shader complexity may increase for the artist to hook up. But this is all dependant on the specific shader pipeline of your engine.
The character modeller then makes the needed LOD's for that character.
The rigger or Tech character artist, sets up all the final rigs and controls. Usually sets up the collision and all the technical gizmos that are needed like Sockets and code controls.
NB: Next gen machines may allow more bones per chunk. So this may add more time in the rigging phase to set up.
Then this is passed over to the animator for animation. Blends, animations and all that.
Then the animations need to be hooked up to the game design and code. So we can say when SPACE BAR is pressed, character plays Jump animation. ect ect.
NB: With more bones per chunk, Animators may need to spend longer amounts of time animating (Yes there is mo-cap which has its own stupid crap to deal with. cleaning up keyframes.. bleh. Also it can't be used for lets say, a crazy squid monster of doom). Though I am not an animator, so I'm only talking from observation.)
Then the character is finally in the game.
Wow finally its in the game! Oh shit... what happens when, for example, you show this to the publisher and he decides you character needs to wear a cloak? Then you have to dismantle the pipeline back to the modelling phase to add the cloak. the rigger/character tech artist then needs to re bake weights and add the new bones and re rig the bastard. Then the animator needs to add all these cloak animations to his existing animations.
One fuckup or one change can bring the whole pipeline to its knees. The more complex it is the harder the pipeline crashes.
Do this times X number of characters in a game. It can add up.
Now environments that's a different story. You usually only need to do up to the collision part. (Unless you had to make a complex prop with animation of course) We need to make collision meshes (soooo boring) and try to find new ways of saving drawcalls while adding more to a scene. Environments of course need a variety of shaders for different materials and different shaders on the one object also means more drawcalls. For example a tree will need two shaders because wood behaves differently than leaves. Its a discipline quite different from characters but takes just as long because the basic process is the same with the baking and the source models.
NB: Artists are generally always wanting to do better than their last project. They always try to push push push. You give them extra resources they will take it, fill it and demand more. That's just how we are. Milk every drop. If a frame rate is shitty in a game its because the artists pushed too hard expecting code to keep up.
For reference I have been both a character and environment artist for a next gen single player game utilising unreal engine 3 and also a next gen MMO. The pipeline I described was what I used when making characters for the single player unreal3 game.
Wow! They should hire you man, you really seem to know your stuff.Learn
To
I pretty much guarantee that Sony and MS would rather ride out the generation as long as possible, especially because HD game development is not as pricey as when the generation first started, but still a bit expensive. There are some developers already having problems with HD consoles. However, the market wants a push onto the next generation and that's when it gets expensive again.
Wow! They should hire you man, you really seem to know your stuff.
Really? I know he's right FFS! It was more that he was pointing out the blatantly obvious, as if dev's don't know the importance of budgeting.Funnily enough he's right. Any dev that sees costs double next gen isn't budgeting properly. Budgets will expand as they do with any new generation but only a poorly budgeted project will see a doubling of costs.
Normal mapping relies on having polygons in the right places, with the right smoothing groups, the right UV seams in the right areas, with the right projection cage, all these little things. It's a pain to have to massage and tweak these assets because our limited performance budgets this generation make every polygon face count so we spend a lot of time noodling to fix common artifacts. Many of these artifacts (rippling normals, seams, projection errors) are ameliorated by simply throwing more vertices at the problem. Also, things like facial deformation and proper loop and edge flow of a 3d model of a head get a lot better if you have the verts to spend to make everything perfect.
...
IMO costs are always going to increase, unless developers make a conscious decision not to keep up with Joneses and move diagonally ahead instead of straight ahead. For example if you look at Journey, the focus there is on sand and clothing simulation, scene construction and cinematography (for lack of a better term), image quality, etc. It's not on normal mapping or super high polygon hands. It's like 2001 to Transformers - the former has superior shots and scene composition, the latter throws a bunch of technically advanced shit all over the screen. If you want to out-Transformers Transformers you have to spend more, because spending on effects is the essence of Transformers. That doesn't mean you can't make a better or even better looking movie.
Similarly if you want to create a "AAAA" game that's better than rival AAA games that largely boils down to spending more on production. If that is your goal then costs aren't going to go down almost be definition.
I feel like everything you said is right, yet your argument as a whole is wrong.
New hardware is not just going to give you bigger textures and higher vertex counts, it's also going to give you new techniques. A lot of the time spent this gen is in making high-res models so you can bake normal maps - before normal maps existed this wasn't an issue.
New hardware is going to create a new breed of normal-maps. Maybe now you'll need to create maps for parallax on every surface, or maps for a different lighting model that is closer to BRDF. Maybe you'll need to do something to take advantage of things like geometry shaders and other DX 10 & 11 features. Maybe megatextures become the norm and now you need to uniquely texture everything. Maybe you'll be painting a texture that represents the rigidity of cloth.
I cant say for sure what those things are, but I'm relatively certain they are going to exist. Smart programmers will come up with new rendering (and other) techniques that require new inputs. This is why every engine created that supposedly scales forward has not actually scaled forward - if you take an engine from 10 years ago that can scale up it doesn't look as good as an engine from today because more has changed than just hardware power.
It's quite possible that the tasks you do today might get a bit faster, but it's also highly probable that you'll have new tasks.
---
IMO costs are always going to increase, unless developers make a conscious decision not to keep up with Joneses and move diagonally ahead instead of straight ahead. For example if you look at Journey, the focus there is on sand and clothing simulation, scene construction and cinematography (for lack of a better term), image quality, etc. It's not on normal mapping or super high polygon hands. It's like 2001 to Transformers - the former has superior shots and scene composition, the latter throws a bunch of technically advanced shit all over the screen. If you want to out-Transformers Transformers you have to spend more, because spending on effects is the essence of Transformers. That doesn't mean you can't make a better or even better looking movie.
Similarly if you want to create a "AAAA" game that's better than rival AAA games that largely boils down to spending more on production. If that is your goal then costs aren't going to go down almost by definition.
Movie production costs haven't gone down, even though CGI has replaced practical effects and new tools and processing power have made CGI theoretically easier and cheaper over time. I'm sure every time some new CGI tool comes out somebody says "this will probably bring down costs" but that has never actually happened.
I'm going to quote this the next time someone suggests that significantly higher development costs are unrelated to more powerful hardware.
Great job breaking this down for us.
Why hasn't every PC dev gone under then? I mean if it's a matter of higher poly models and high-res textures, the PC has had those for years...just check the Witcher 2 360/PC comparison thread for reams and reams of gleefully-posted evidence