Note: This post is mainly focused on the cost of content (art) asset creation. There are obviously other significant development costs (engine technology, developer tools, staffing to create such) in addition to artists and designers, but my interest here was to "think out loud" about potential future product costs of content creation on future consoles.
One of the major factors in this generation of software development has been the increase in development budgets. A number of factors contributed to this issue but it seems most agree that content creation and iteration have been the major factor in this equation. Content is expensive to create and the current platforms require (a) more unique assets and (b) higher quality assets to compete.
The question is if this curve will flatten out for typical titles next generation.
I think there is some evidence we won't see development budgets explode again, at least for savvy dev studios. Here are some reasons I think AAA titles can be made without a significant bump in current budgets while still obtaining significant increases in expected graphical fidelity.
1. Source Art. Most current generation games use normal maps and parallax maps for added detail in game assets. A 10,000 poly character will often have a 1,000,000 source model used to generate the normal maps. As Epic has noted, a 1,000,000 poly game level will be derived from a 200,000,000 poly source, the difference being the normal maps "storing" the difference in poly detail.
Is there a reason that this generations 1M poly character source model needs to become a 10M poly source model? If next gen sees a jump from 10M polys + normal maps to 30M polys + parallax (displacement?) maps with tesselation is there a practical need to bump the source model up? The diminishing returns in the consoles (discussed in the prediction thread) have some bearing here on the source side as well. What is there to be gained in a practical sense in increasing the detail with a target resolution of 1080p?
It seems to me, due to the use of normal maps, the source art is already significantly detailed and the issue is less about the quality of the source art but more the abilities of the hardware to replicate the quality.
2. Fidelity and Technology. If my arguement above stands, i.e. source art quality is high enough in many cases already to be re-purposed for next gen, it would seem logical that the conjunction of high quality game assets from the same source models in conjunction with better technology could make a visual leap with minimal development cost increase.
Take Gears of War 2. How much additional development time is going to be required for the artists if the game were running on hardware that was capable of higher quality textures, 1080p, 4xMSAA, 8xAF, and Epic was able to deploy dynamic shadows, dynamic lighting, and a better GI hack? All of these things are possible using the current source art. In some ways realtime technologies can help with iteration.
Better rendering techniques and higher fidelity rendering in conjunction with higher quality game assets from current generation source art assets could provide a significant visual punch without a steep increase in source art creation costs. At least in theory.
3. Procedural Tools. There are a number of middleware tools for procedural content generation (clouds, vegitation, etc). Companies like Crytek and Ubisoft (Far Cry 2) have shown how these tools can be utilized to quickly create unique locals and Epic has shown in UE3 the use of such tools to quickly create unique geometry. idTech5 goes another direction with essentially a giant flat canvas where "stamping" allows for the quick creation of unique detail throughout the world. Many companies use procedural tools to create unique looking NPCs instead of individually modeling every detail of a character.
While procedurals tools aren't a one size fits all solution it seems current tools (especially retime tools focused on iteration) speed up content creation. New consoles will still have limits (memory, rendering power, importantly processing power and efficency) but the issue may be one about "how much grass do you want?" and not "can we do grass or not?" Tools that allow artists to quickly deploy basic assets in realtime should contunie to speed up development on the art side. How far are we from seeing tools that create roads and buildings?
A hypothetical "procedural building maker" could simply prompt the artists:
Style? (Modern, Victorian, Industrial, Cottage, Midevil...)
Age? (Years)
Condition? (Great, typical, poor maintenance)
Siding type?
Color?
Roof type?
Stories?
Windows?
Doors?
Accessories? (Porch, Garage, basement, etc)
After creating the base unit the artist could then move around the doors and windows via a WYSIWYG, add shutters, etc. Decals (vandalism?) and unique subtleties could be added to each building in a final pass. Rooms could be populated with a "fill" tool with assets of set categories and re-arranged. Entire neighborhoods, even cities, could be deployed quickly through such tools. Just as Far Cry 2 allows you to create entire forests on consoles (!) with unique terrain and foliage is it not plausable that these sort of tools will be extended to allow quick iteration of unique, high quality asset development?
We have already seen in Far Cry 2 how a developer in 1 hour built a multicamp map with connecting roads, vegitation, etc, set the time of day, wind, vehicle placement etc all within 1 hour. In past generation creating such would have taken a team many man hours. Yes, the tools required significant amounts of time to make and the asset library also was a sizable investment. Yet we are seeing with Dunia that ubisoft is repurposing the foliage engine. The large up front cost of developing procedural tools isn't insignificant, but the ability to repurpose these tools and assets to expedite unique content creation seems like a direction to increase quality while controlling development costs.
Bit-tech ran an article a while back on procedural "aging" of materials through shaders. This is another unique concept where unique high quality models can be created quickly without requiring the artists direct input on every pixel. The example was a bathroom that aged and was damaged, all changed through shaders.
4. Art. Two trajectories in regards to art could help keep costs low. First is "realistic" games. With the continued creation of high quality assets intended to be utilized for games with a "realistic" look there should be a growing library of assets at publishers for utilization. While not every asset can be re-used, minimally using a standard library as a starting point to get a project moving and customizing as the project matures seems plausible. A unique feel could also be added not so much through the recreation of every asset but by the use of rendering techniques that give a unique look as well as shaders that give objects a unique look and wear to match the tone of the art direction. Likewise more "artistic" games create opportunities to create high quality, unique looking games that put more emphasis on style and unique rendering tricks above brute asset creation. Looking at CGI and comic books there is a large variety of art concepts that can be deployed that shift budgets away from "more and better". Artists may be in a position to create unique worlds that match their artistic vision that significantly underutilize the hardware (and are affordable to create) yet have a significant visual punch.
My general point is that I see ways for savvy studios to create cutting edge software without breaking the bank next gen. This may create some limits of the type of game developers create and this doesn't begin to address the difficulties developers face in deploying new technology (animation, AI, physics, all these tools to help artists and designers, etc aren't free or easy). Likewise there will always be the high profile games that throw caution to the wind or "crunch" products that abandon fiscal sense to get a product out on time, but it would appear that developers at least have options next generation to continue improving the visual quality of games without doubling, tripling, or even quadrupaling their current budgets. Which also makes sense with diminishing returns--the visual return for a 200% increase in mesh fidelity for physics or 5x improvement in poly count is less appearant there is a logical move away from investing in something many consumers cannot see and therefor don't value. Bigger, unique, and more interactive worlds; refined and unique gameplay experiences; deep featuresets with varied experiences; stories... and so forth have a direct effect on consumers. We have seen with a number of high profile, high selling titles that the development focus has less been on strict visuals but on user experience. If consumers are choosing these titles over technically and visually superior products based on art preference and user experience it would appear to be a solid guide to next gen software development where iteration is a primary focus of development. Hence limiting the costs of asset development and smart (re-)utilization of assets and procedural tools may be avenues which we see dev budgets remain flat while the general quality of games continues to improve.
So do developers feel there is some truth to the above, or are there some fundamental flaws in understanding the direction, and cost associations, with deploying said approaches. My hunch is there are studios who will find ways to raise the graphical bar while keeping budgets similar to what they have today. Disagree?
One of the major factors in this generation of software development has been the increase in development budgets. A number of factors contributed to this issue but it seems most agree that content creation and iteration have been the major factor in this equation. Content is expensive to create and the current platforms require (a) more unique assets and (b) higher quality assets to compete.
The question is if this curve will flatten out for typical titles next generation.
I think there is some evidence we won't see development budgets explode again, at least for savvy dev studios. Here are some reasons I think AAA titles can be made without a significant bump in current budgets while still obtaining significant increases in expected graphical fidelity.
1. Source Art. Most current generation games use normal maps and parallax maps for added detail in game assets. A 10,000 poly character will often have a 1,000,000 source model used to generate the normal maps. As Epic has noted, a 1,000,000 poly game level will be derived from a 200,000,000 poly source, the difference being the normal maps "storing" the difference in poly detail.
Is there a reason that this generations 1M poly character source model needs to become a 10M poly source model? If next gen sees a jump from 10M polys + normal maps to 30M polys + parallax (displacement?) maps with tesselation is there a practical need to bump the source model up? The diminishing returns in the consoles (discussed in the prediction thread) have some bearing here on the source side as well. What is there to be gained in a practical sense in increasing the detail with a target resolution of 1080p?
It seems to me, due to the use of normal maps, the source art is already significantly detailed and the issue is less about the quality of the source art but more the abilities of the hardware to replicate the quality.
2. Fidelity and Technology. If my arguement above stands, i.e. source art quality is high enough in many cases already to be re-purposed for next gen, it would seem logical that the conjunction of high quality game assets from the same source models in conjunction with better technology could make a visual leap with minimal development cost increase.
Take Gears of War 2. How much additional development time is going to be required for the artists if the game were running on hardware that was capable of higher quality textures, 1080p, 4xMSAA, 8xAF, and Epic was able to deploy dynamic shadows, dynamic lighting, and a better GI hack? All of these things are possible using the current source art. In some ways realtime technologies can help with iteration.
Better rendering techniques and higher fidelity rendering in conjunction with higher quality game assets from current generation source art assets could provide a significant visual punch without a steep increase in source art creation costs. At least in theory.
3. Procedural Tools. There are a number of middleware tools for procedural content generation (clouds, vegitation, etc). Companies like Crytek and Ubisoft (Far Cry 2) have shown how these tools can be utilized to quickly create unique locals and Epic has shown in UE3 the use of such tools to quickly create unique geometry. idTech5 goes another direction with essentially a giant flat canvas where "stamping" allows for the quick creation of unique detail throughout the world. Many companies use procedural tools to create unique looking NPCs instead of individually modeling every detail of a character.
While procedurals tools aren't a one size fits all solution it seems current tools (especially retime tools focused on iteration) speed up content creation. New consoles will still have limits (memory, rendering power, importantly processing power and efficency) but the issue may be one about "how much grass do you want?" and not "can we do grass or not?" Tools that allow artists to quickly deploy basic assets in realtime should contunie to speed up development on the art side. How far are we from seeing tools that create roads and buildings?
A hypothetical "procedural building maker" could simply prompt the artists:
Style? (Modern, Victorian, Industrial, Cottage, Midevil...)
Age? (Years)
Condition? (Great, typical, poor maintenance)
Siding type?
Color?
Roof type?
Stories?
Windows?
Doors?
Accessories? (Porch, Garage, basement, etc)
After creating the base unit the artist could then move around the doors and windows via a WYSIWYG, add shutters, etc. Decals (vandalism?) and unique subtleties could be added to each building in a final pass. Rooms could be populated with a "fill" tool with assets of set categories and re-arranged. Entire neighborhoods, even cities, could be deployed quickly through such tools. Just as Far Cry 2 allows you to create entire forests on consoles (!) with unique terrain and foliage is it not plausable that these sort of tools will be extended to allow quick iteration of unique, high quality asset development?
We have already seen in Far Cry 2 how a developer in 1 hour built a multicamp map with connecting roads, vegitation, etc, set the time of day, wind, vehicle placement etc all within 1 hour. In past generation creating such would have taken a team many man hours. Yes, the tools required significant amounts of time to make and the asset library also was a sizable investment. Yet we are seeing with Dunia that ubisoft is repurposing the foliage engine. The large up front cost of developing procedural tools isn't insignificant, but the ability to repurpose these tools and assets to expedite unique content creation seems like a direction to increase quality while controlling development costs.
Bit-tech ran an article a while back on procedural "aging" of materials through shaders. This is another unique concept where unique high quality models can be created quickly without requiring the artists direct input on every pixel. The example was a bathroom that aged and was damaged, all changed through shaders.
4. Art. Two trajectories in regards to art could help keep costs low. First is "realistic" games. With the continued creation of high quality assets intended to be utilized for games with a "realistic" look there should be a growing library of assets at publishers for utilization. While not every asset can be re-used, minimally using a standard library as a starting point to get a project moving and customizing as the project matures seems plausible. A unique feel could also be added not so much through the recreation of every asset but by the use of rendering techniques that give a unique look as well as shaders that give objects a unique look and wear to match the tone of the art direction. Likewise more "artistic" games create opportunities to create high quality, unique looking games that put more emphasis on style and unique rendering tricks above brute asset creation. Looking at CGI and comic books there is a large variety of art concepts that can be deployed that shift budgets away from "more and better". Artists may be in a position to create unique worlds that match their artistic vision that significantly underutilize the hardware (and are affordable to create) yet have a significant visual punch.
My general point is that I see ways for savvy studios to create cutting edge software without breaking the bank next gen. This may create some limits of the type of game developers create and this doesn't begin to address the difficulties developers face in deploying new technology (animation, AI, physics, all these tools to help artists and designers, etc aren't free or easy). Likewise there will always be the high profile games that throw caution to the wind or "crunch" products that abandon fiscal sense to get a product out on time, but it would appear that developers at least have options next generation to continue improving the visual quality of games without doubling, tripling, or even quadrupaling their current budgets. Which also makes sense with diminishing returns--the visual return for a 200% increase in mesh fidelity for physics or 5x improvement in poly count is less appearant there is a logical move away from investing in something many consumers cannot see and therefor don't value. Bigger, unique, and more interactive worlds; refined and unique gameplay experiences; deep featuresets with varied experiences; stories... and so forth have a direct effect on consumers. We have seen with a number of high profile, high selling titles that the development focus has less been on strict visuals but on user experience. If consumers are choosing these titles over technically and visually superior products based on art preference and user experience it would appear to be a solid guide to next gen software development where iteration is a primary focus of development. Hence limiting the costs of asset development and smart (re-)utilization of assets and procedural tools may be avenues which we see dev budgets remain flat while the general quality of games continues to improve.
So do developers feel there is some truth to the above, or are there some fundamental flaws in understanding the direction, and cost associations, with deploying said approaches. My hunch is there are studios who will find ways to raise the graphical bar while keeping budgets similar to what they have today. Disagree?
