Hey Kevin, amazing work, really appreciate you posting this here. You Epic guys are really the epitome of next-gen
I do have a question:
Don´t you ever think that you´re putting too much detail in these highpoly meshes? So much that it will never all translate to the normalmaps? I´m just wondering because there´s a point where one can just create too much detail that it isn´t worth the time anymore. But maybe that´s why you are the best, doing all that detail and still being fast at it.
Do I put too much detail in sometimes, yes. The above section of the Hospital Facade suffers from my trying to push too much into the texture, I didn't think it through enough, I should have dropped the leaves from the door frame, I should have dropped the inner door blocks.
I didn't prioritize the door section with enough resolution, I placed an equal amount on the whole asset when I should have realized I could have scaled the res down a bit on the side and upper section a bit more to allow for enough of the details to come through.
That's the one piece in the set that I really wish I had done things differently, I always felt I dropped the ball a little on this one and I'm trying harder to be more restrained in general these days. I hope I am becoming more balanced as I go forward.
Hiindsight helps, its part of why I'm posting also, it helps confirm or deny the things I see in hindsight along with the rest of you.
I can't help but feel this sort of workflow (adding high density meshes in to max, straight out of zbrush/mudbox) isn't something the average user will be able to do in a manageable state. I cringed when I saw the step of your bricks that had you bringing it back in to Max and instancing it around. Generally, Max doesn't handle that high of a triangle count. (So - system specs?)
With that, could these bricks not be projected down to a simple plane, normals baked, then manipulate said plane/geo as you have in your later descriptions and achieve the same look?
omg omg omg do i have birthday again xD? thanks for sharing Kevin. I should really take a deeper look into that "grid is master" stuff. that lantern looks crazy hehe
I can't help but feel this sort of workflow (adding high density meshes in to max, straight out of zbrush/mudbox) isn't something the average user will be able to do in a manageable state. I cringed when I saw the step of your bricks that had you bringing it back in to Max and instancing it around. Generally, Max doesn't handle that high of a triangle count. (So - system specs?)
With that, could these bricks not be projected down to a simple plane, normals baked, then manipulate said plane/geo as you have in your later descriptions and achieve the same look?
Yeah, I have to agree, my machine would bomb out on me if I tried that . I think bending a flat plane probably would have issues. There's always a difference between rendering a proper high poly, and cutting corners. The mesh catches things in different ways when projected. Also, you wouldn't be able to have tapered sections like he does with that dome very easily.
Great work though, I love seeing high poly stuff .
I still would like to think there's more hardware efficient ways of doing what Kevin's done.
Also, Kev, this picture here has a shadow error on the bottom left side of the dome.
While we're on the dome - I take it this is just a beauty shot of the highpoly work, ya? It wasn't built like this when baked down, right? It's a pretty modular design, so I'd imagine 1/6th of the dome's sections would be what you actually would use for baking, ya?
I ask because this is where I hit a small bump whenever doing highpoly. Do I flesh it out completely even though I'm only using small sections of it for baking while the rest is just duplicated for tart on my end? (For the possibility of beauty renders later on). This is assuming its not all 1 large unique unwrap with unique texture detailing everywhere.
thanks, so much Kevin for taking the time out, to share your work with us. There are not a lot of professionals out there who share the information they have learned over the years. Everything you post is so insightful and inspirational.
so how do you go about figuring the sizes of modular pieces? Do you have any preset ideas like a wall should be x units tall by x units high and corner pieces should be a certain size, etc... ?
I can't help but feel this sort of workflow (adding high density meshes in to max, straight out of zbrush/mudbox) isn't something the average user will be able to do in a manageable state. I cringed when I saw the step of your bricks that had you bringing it back in to Max and instancing it around. Generally, Max doesn't handle that high of a triangle count. (So - system specs?)
With that, could these bricks not be projected down to a simple plane, normals baked, then manipulate said plane/geo as you have in your later descriptions and achieve the same look?
Well the cool thing is, that all of the concept he uses easily apply to pretty much whatever level of detail your pc can handle. If you dont have 16 gigs, you can pretty easily just skip the zbrush step on the bricks for example, and do that later in PS or just not go to that level of sub-d, you can get some nice shape out of a brick like that without a whole lot of geometry. The main thing is the quality of the forms anyway, the little texture detail is moot as far as i'm concerned. Model it or paint it, its the design and the main forms that will sell your model.
I still would like to think there's more hardware efficient ways of doing what Kevin's done.
While we're on the dome - I take it this is just a beauty shot of the highpoly work, ya? It wasn't built like this when baked down, right? It's a pretty modular design, so I'd imagine 1/6th of the dome's sections would be what you actually would use for baking, ya?
I ask because this is where I hit a small bump whenever doing highpoly. Do I flesh it out completely even though I'm only using small sections of it for baking while the rest is just duplicated for tart on my end? (For the possibility of beauty renders later on). This is assuming its not all 1 large unique unwrap with unique texture detailing everywhere.
For myself i find that a good workflow is to block out the entire thing in highpoly, even if only a section of it will get baked. Its hard to get a feel for the entire thing when you can only see a fraction of it, and for something like this its really easy just to instance it or whatever. I think in this case it works really well.
Now if you're doing say, an entire bulding with mostly tiling textures, and a few unique, baked assets, it makes less sense to do more than the lowpoly blockout for the entire thing. Again you could pretty easily set up instances in this case, to get a better overall feel for it, but it may not be worthwhile.
I've been rendering up and mocking out these tutorial images with a 2gb machine on windows xp that I got shortly after Gears1 shipped. It's the same kind of machine I used to make the hospital set, in fact I remember that my home pc back then was more powerful than my work machine until I got on to the locust set.
The workflow I've been showing is easily manageable with proper planning. I am trying to show this.
Theres 8 bricks, each brick is around 60,000 tri's, they are each instanced about 8-10 times, theres a floating plane of wobbly polys in a different material to act as the grout inbetween the bricks.
When I process that, its just to a flat plane, everything else from the scene is deleted, max is given a fresh restart before I process.
It's the flat plane that I then shell out, unwrap the sides, unwrap the 2ndry lightmap uv's then then subdivide and path deform WSM along any spline shape on the ground I choose to draw.
I applied a bend modifier in the shot I showed to communicate that the bricks in the scene are modifiable in any manner you want because they were there in the scene I was manipulating and the lows were in another and I was lazy. So I probably misdirected you by demonstrating with the hi instead of the low, sorry about that.
Anyway, in each case where scenes have an object that is repeated, like in the dome shot, each of those forms are only modelled once and each are processed as seperatly as required, often I have to break files up into seperate rendering files because there are, as people are pointing out, limitations to how many zbrush chunks of tris that max will process with a skylight and light tracer on in order to get a nice lightmap out too as a head start for the texture guys.
So it comes back to pre planning again, all this stuff is manageable as long as you think ahead.
Initially I didn't know this, I tried too often at first to do it all in one block, rather than taking the extra time to think of each element of a design as a whole element. If a trim like the grilled piece in the hospital set has a rectangular block, I initially model it like one and keep it as a solid element. Each longer element like the grilled central section is a decent sized whole block and needs to stay that way so I carved grooved cuts equally subdviding the piece and work that into the design so I can treat the main element as a collection of near identical smaller elements which are nice and portable in Zbrush via subtools once I take the polycount up much higher.
Doing this, makes it all easier on the back end when I bring them back into max.
It's not hard though, it's just tedious, but theres lots of parts of the process that you can automate. The Polycruncher plugin has a great Batch Optimize option which will mindlessly churn through whatever directory you point it toward, while you sit back and read a book as your pc is busy.
Once it's done with that you can use Blue's batch importer to bring the optimized and named chunks into max for you, ready to apply a material to, sanity check each piece and process.
While you process, you read a book again.
I still rely on the idea that artists are lazy and patience is a virtue.
It's a longwinded workflow to consider at first, but it gets to be something you can do on autopilot for a large percentage of it, leaving only the real problem solving part to be the creative aspects.
Other questions...
To model the whole thing or not to model the whole thing?
Well I tend to model out the whole thing, but I very quickly cut whatever it is down to 1/4 or 1/8 of what it is by instancing around a pivot point on the grid so that I am not only cutting down on the drain on max and cutting down on how much I need to model, but I'm also anticipating how I'm going to unwrap in an efficient way.
Usually I start the lowpoly halfway through doing the hipoly, I unwrap the low before I finish the hipoly so I know if there is space left unused because of unusual shape, if there is I model some kind of extra attachment for the design to fit in that space.... why waste the space?
This is the point where you snicker at the Scottish penny (uvw space) pinching stereotype probably.
It's this sort of back end stuff that makes the front end more robust, you can't see it in the game, but its making the game run faster, the reward for it is knowing you did it. You did as right as you could.
And a year later you look back and wonder why you did it so dumb. I'm not sure as there are any really good tricks to get through it easier, theres only one great trick I know which is to just keep doing it all the time until you make it easier for yourself to do it.
Yes I see I did that render wrong, I'm not much use with rendering, I was worse back then when that render was made. I gave up on taking the time to render new stuff and started modifying the renders I did at the time onto a grey background. You keeping me sober here Adam ?
Well, thanks again. I appreciate the encouragement, I think I answered most of the questions or near enough it in a round about way right?
I think the overall arching message here that people can take away is that with proper planning at the beginning of a project, whether its an environment or character, will save you a lot of work in the long wrong. So spending the time early to think things out, which may or may not mean jumping right in an pushing polys as I am sure some will do, is an efficient and effective way of working.
Right on Kevin. Thanks for writing that out. And, if this thread is any indication, the proof is in the pudding.
wicked write up man, great insight into your workflow. I know personally I over think and over complicate things a lot of the time and your advice is a great help on the key points to think about beforehand and and keeping it as straightforward as possible while working through the piece.
Excellent explanation Kevin. Sometimes planning hurts my mind because planning stages can go in all directions leaving me confused because there are so many routes to take.
thanks for that Kev, a lot of information provided by your good honest self there :poly121:
I have a question if anyone can answer, does working on the grid apply to all modelling for the different engines. Sorry if this is a naive question , I come from a archvis background and desperately need to know more about the modelling workflow for games.
I have a question if anyone can answer, does working on the grid apply to all modelling for the different engines. Sorry if this is a naive question , I come from a archvis background and desperately need to know more about the modelling workflow for games.
Working on a grid is more important in some engines, like UE3 than it is in others but ultimately its good practice to make sure you stick to it whatever your doing. It just makes constructing modular peices and environments far far easier if everything snaps neatly into place.
thanks for the massive helpful write-ups kev, started a new folder for all of this on my desktop :P
i think you talked about this in the older thread but when doing the high poly for somthing like the flatter concrete walls or whatever, how do handle tileability and adding the different levels of detail over that?
Tielable alpha projected in zbrush and then detail added over sound about right?
Anyway, thanks for the awsome new art man, gotta go steal my friends xbox to play gow2 some time :P
Working on a grid is more important in some engines, like UE3 than it is in others but ultimately its good practice to make sure you stick to it whatever your doing. It just makes constructing modular peices and environments far far easier if everything snaps neatly into place.
what defines the size of this grid.?
In archvis, our 3dsmax units would be mm, and my display unit scale would be in metres. I'm attempting to model 'a church' typical fare for my portfolio, and Ive adopted mm and metres, (since I'm used to it, and its real world scale) Ive kept my model to the max grid as close as poss, Should I be modelling to different units, or are these units defined by the Unreal editor ?? for example.
i think I found a thread or tutorial on modular modelling to the grid, but am trying to find it again with no luck. If anyone knows where it is please direct me. thanks for any help guys. I'll probably be doing a lot of this asking.. spent enough years at uni, teaching myself now (but then I did at uni too)..
Thanks for sharing Kevin. I take it that was a glimpse into your upcoming Eat3D tutorial "the grid is god"?
You have an Eat3D tutorial DVD coming??? That would be great! Will definitely purchase that if in fact there is one.
Thank you again for sharing your work flow with us and taking the time.
There is already a lot of user created content being made for UT3 and using UE3, but I wonder if there were some DVDs or tutorials like this from people from within Epic pertaining to work flow, that it would spur on more people to create content. I know that is a bit ridiculous to expect any of them including you to have time to even participate in something like that on top of working and family. It feels like there was far more creation going on within the community before UE3 and I always wonder if that is due to the work flow / process of creation being not that it is necessarily more complex but appearing to be more complex in the minds of people than it really is. It definitely entails more work / time and like you said pre-planning and forethought. I believe the UE3 books from the 3D Buzz people are still on their way, but not until June and July respectively. Eat 3D has definitely been a great source of information and insight...Thank you Riki.
Again thank you for taking the time to put this together. It is appreciated.
This is such great stuff. The design are borderline absurd, but in a good way. You can focus on an ever smaller section of a prop, and there's always more detail to be found.
You do have a lot of nifty tricks up your sleeve Kevin, but really, the biggest thing I've taken from your work samples is that you have to work your ASS off.
How do you drink your coffee? Black, served in a cup with volute handles, covered with desert flowers, and star trek pipes running along the bottom?
jk
Replies
I do have a question:
Don´t you ever think that you´re putting too much detail in these highpoly meshes? So much that it will never all translate to the normalmaps? I´m just wondering because there´s a point where one can just create too much detail that it isn´t worth the time anymore. But maybe that´s why you are the best, doing all that detail and still being fast at it.
1) did you only work on the high-poly assets or did you also translate these into low-poly models? and if so can you show the assets?
2) how do you render your images they look really nice and i would love to have some sexy renders like that in my poertfolio.
And another straight pimp -
Do I put too much detail in sometimes, yes. The above section of the Hospital Facade suffers from my trying to push too much into the texture, I didn't think it through enough, I should have dropped the leaves from the door frame, I should have dropped the inner door blocks.
I didn't prioritize the door section with enough resolution, I placed an equal amount on the whole asset when I should have realized I could have scaled the res down a bit on the side and upper section a bit more to allow for enough of the details to come through.
That's the one piece in the set that I really wish I had done things differently, I always felt I dropped the ball a little on this one and I'm trying harder to be more restrained in general these days. I hope I am becoming more balanced as I go forward.
Hiindsight helps, its part of why I'm posting also, it helps confirm or deny the things I see in hindsight along with the rest of you.
With that, could these bricks not be projected down to a simple plane, normals baked, then manipulate said plane/geo as you have in your later descriptions and achieve the same look?
Thanks for the workflow post, Kevin, it's very informative and sensible. I'm sure a lot of people can learn a ton of stuff from these assets
That dome is really nicely done too, probably my favourite of this new bunch. The lighting / colour setup you have is working well too.
Awesome work as always buddy.
Yeah, I have to agree, my machine would bomb out on me if I tried that
Great work though, I love seeing high poly stuff
Also, Kev, this picture here has a shadow error on the bottom left side of the dome.
While we're on the dome - I take it this is just a beauty shot of the highpoly work, ya? It wasn't built like this when baked down, right? It's a pretty modular design, so I'd imagine 1/6th of the dome's sections would be what you actually would use for baking, ya?
I ask because this is where I hit a small bump whenever doing highpoly. Do I flesh it out completely even though I'm only using small sections of it for baking while the rest is just duplicated for tart on my end? (For the possibility of beauty renders later on). This is assuming its not all 1 large unique unwrap with unique texture detailing everywhere.
love the breakdown very much
@adambrome: i believe kevin said in his other thread that he has 16gigs of ram.
and badass kev! i admire how you take the time to post your workflow and explain things. I hope one day I am allowed to do that
EDIT: This sounds like cockery on my part but its a genuine question, I promise
It would be the same way.....but it would take 5 days to import the brick. ;D
Thanks for sharing your work flow as well.
Needs more Raam!
Well the cool thing is, that all of the concept he uses easily apply to pretty much whatever level of detail your pc can handle. If you dont have 16 gigs, you can pretty easily just skip the zbrush step on the bricks for example, and do that later in PS or just not go to that level of sub-d, you can get some nice shape out of a brick like that without a whole lot of geometry. The main thing is the quality of the forms anyway, the little texture detail is moot as far as i'm concerned. Model it or paint it, its the design and the main forms that will sell your model.
For myself i find that a good workflow is to block out the entire thing in highpoly, even if only a section of it will get baked. Its hard to get a feel for the entire thing when you can only see a fraction of it, and for something like this its really easy just to instance it or whatever. I think in this case it works really well.
Now if you're doing say, an entire bulding with mostly tiling textures, and a few unique, baked assets, it makes less sense to do more than the lowpoly blockout for the entire thing. Again you could pretty easily set up instances in this case, to get a better overall feel for it, but it may not be worthwhile.
The workflow I've been showing is easily manageable with proper planning. I am trying to show this.
Theres 8 bricks, each brick is around 60,000 tri's, they are each instanced about 8-10 times, theres a floating plane of wobbly polys in a different material to act as the grout inbetween the bricks.
When I process that, its just to a flat plane, everything else from the scene is deleted, max is given a fresh restart before I process.
It's the flat plane that I then shell out, unwrap the sides, unwrap the 2ndry lightmap uv's then then subdivide and path deform WSM along any spline shape on the ground I choose to draw.
I applied a bend modifier in the shot I showed to communicate that the bricks in the scene are modifiable in any manner you want because they were there in the scene I was manipulating and the lows were in another and I was lazy. So I probably misdirected you by demonstrating with the hi instead of the low, sorry about that.
Anyway, in each case where scenes have an object that is repeated, like in the dome shot, each of those forms are only modelled once and each are processed as seperatly as required, often I have to break files up into seperate rendering files because there are, as people are pointing out, limitations to how many zbrush chunks of tris that max will process with a skylight and light tracer on in order to get a nice lightmap out too as a head start for the texture guys.
So it comes back to pre planning again, all this stuff is manageable as long as you think ahead.
Initially I didn't know this, I tried too often at first to do it all in one block, rather than taking the extra time to think of each element of a design as a whole element. If a trim like the grilled piece in the hospital set has a rectangular block, I initially model it like one and keep it as a solid element. Each longer element like the grilled central section is a decent sized whole block and needs to stay that way so I carved grooved cuts equally subdviding the piece and work that into the design so I can treat the main element as a collection of near identical smaller elements which are nice and portable in Zbrush via subtools once I take the polycount up much higher.
Doing this, makes it all easier on the back end when I bring them back into max.
It's not hard though, it's just tedious, but theres lots of parts of the process that you can automate. The Polycruncher plugin has a great Batch Optimize option which will mindlessly churn through whatever directory you point it toward, while you sit back and read a book as your pc is busy.
Once it's done with that you can use Blue's batch importer to bring the optimized and named chunks into max for you, ready to apply a material to, sanity check each piece and process.
While you process, you read a book again.
I still rely on the idea that artists are lazy and patience is a virtue.
It's a longwinded workflow to consider at first, but it gets to be something you can do on autopilot for a large percentage of it, leaving only the real problem solving part to be the creative aspects.
Other questions...
To model the whole thing or not to model the whole thing?
Well I tend to model out the whole thing, but I very quickly cut whatever it is down to 1/4 or 1/8 of what it is by instancing around a pivot point on the grid so that I am not only cutting down on the drain on max and cutting down on how much I need to model, but I'm also anticipating how I'm going to unwrap in an efficient way.
Usually I start the lowpoly halfway through doing the hipoly, I unwrap the low before I finish the hipoly so I know if there is space left unused because of unusual shape, if there is I model some kind of extra attachment for the design to fit in that space.... why waste the space?
This is the point where you snicker at the Scottish penny (uvw space) pinching stereotype probably.
It's this sort of back end stuff that makes the front end more robust, you can't see it in the game, but its making the game run faster, the reward for it is knowing you did it. You did as right as you could.
And a year later you look back and wonder why you did it so dumb. I'm not sure as there are any really good tricks to get through it easier, theres only one great trick I know which is to just keep doing it all the time until you make it easier for yourself to do it.
Yes I see I did that render wrong, I'm not much use with rendering, I was worse back then when that render was made. I gave up on taking the time to render new stuff and started modifying the renders I did at the time onto a grey background. You keeping me sober here Adam ?
Well, thanks again. I appreciate the encouragement, I think I answered most of the questions or near enough it in a round about way right?
Right on Kevin. Thanks for writing that out. And, if this thread is any indication, the proof is in the pudding.
cheers man.
Great work!
I have a question if anyone can answer, does working on the grid apply to all modelling for the different engines. Sorry if this is a naive question , I come from a archvis background and desperately need to know more about the modelling workflow for games.
Working on a grid is more important in some engines, like UE3 than it is in others but ultimately its good practice to make sure you stick to it whatever your doing. It just makes constructing modular peices and environments far far easier if everything snaps neatly into place.
i think you talked about this in the older thread but when doing the high poly for somthing like the flatter concrete walls or whatever, how do handle tileability and adding the different levels of detail over that?
Tielable alpha projected in zbrush and then detail added over sound about right?
Anyway, thanks for the awsome new art man, gotta go steal my friends xbox to play gow2 some time :P
what defines the size of this grid.?
In archvis, our 3dsmax units would be mm, and my display unit scale would be in metres. I'm attempting to model 'a church' typical fare for my portfolio, and Ive adopted mm and metres, (since I'm used to it, and its real world scale) Ive kept my model to the max grid as close as poss, Should I be modelling to different units, or are these units defined by the Unreal editor ?? for example.
i think I found a thread or tutorial on modular modelling to the grid, but am trying to find it again with no luck. If anyone knows where it is please direct me.
You have an Eat3D tutorial DVD coming??? That would be great! Will definitely purchase that if in fact there is one.
Thank you again for sharing your work flow with us and taking the time.
There is already a lot of user created content being made for UT3 and using UE3, but I wonder if there were some DVDs or tutorials like this from people from within Epic pertaining to work flow, that it would spur on more people to create content. I know that is a bit ridiculous to expect any of them including you to have time to even participate in something like that on top of working and family. It feels like there was far more creation going on within the community before UE3 and I always wonder if that is due to the work flow / process of creation being not that it is necessarily more complex but appearing to be more complex in the minds of people than it really is. It definitely entails more work / time and like you said pre-planning and forethought. I believe the UE3 books from the 3D Buzz people are still on their way, but not until June and July respectively. Eat 3D has definitely been a great source of information and insight...Thank you Riki.
Again thank you for taking the time to put this together. It is appreciated.
You do have a lot of nifty tricks up your sleeve Kevin, but really, the biggest thing I've taken from your work samples is that you have to work your ASS off.
How do you drink your coffee? Black, served in a cup with volute handles, covered with desert flowers, and star trek pipes running along the bottom?
jk
Awesome work