3DS Model Organisation and RTT Workflow
polycounter lvl 9
I'd love to know how you guys organise your models for a complex RTT bake.
To put this in context I've found it extremely useful to try and get as much info out of a RTT bake as possible.
Assigning colour coded materials to a model and baking out the diffuse information to create masks is something I find extremely usefull. Its so easy to make these into layer masks in photoshop and they make general painting much neater and creation of spec/gloss maps a million times easier.
It is however something you cant always do. When you get lots of closely packed detail I've found i have to use the 'hit only matching Mat ID's' option to stop bad baking from happening. meaning i cant meaningfully assign different materials to the model anymore.
Now I could explode my meshes but it ruins ambient occlusion calculation.
I found 1 solution was to assign vertex colours to the mesh, sorting the mesh into meaninfull colour coded groups and baking this out in the diffuse channel. This obviously increses the footprint of the model in ram but it works really well.
Another speedbump I've hit is not being able to RTT my whole model at once, my last model was very complex and max just ground to a halt if I told it to bake everything. to this end i had to split the model into groups which i then baked out seprately. This was fine, I added some overlapping areas to the bakes to preserve AO baking and the individual bakes worked well. Comping it all together was a bit of a pain though.
I got round this problem in some small way by rendering as PNG's with alpha and using the max composite node to combine all the baked maps together into 1 tile. It was a little slow to setup but it only had to be done once and it meant I didn't have to keep photoshop open in order to update my normal maps. I think im going to write a script to automate this process, it should be pretty easy to get all the RTT maps from a group of objects and create a comp map of all the different passes.
So how do you approach this sort of thing? Any sage advice for someone trying to streamline their workflow
To put this in context I've found it extremely useful to try and get as much info out of a RTT bake as possible.
Assigning colour coded materials to a model and baking out the diffuse information to create masks is something I find extremely usefull. Its so easy to make these into layer masks in photoshop and they make general painting much neater and creation of spec/gloss maps a million times easier.
It is however something you cant always do. When you get lots of closely packed detail I've found i have to use the 'hit only matching Mat ID's' option to stop bad baking from happening. meaning i cant meaningfully assign different materials to the model anymore.
Now I could explode my meshes but it ruins ambient occlusion calculation.
I found 1 solution was to assign vertex colours to the mesh, sorting the mesh into meaninfull colour coded groups and baking this out in the diffuse channel. This obviously increses the footprint of the model in ram but it works really well.
Another speedbump I've hit is not being able to RTT my whole model at once, my last model was very complex and max just ground to a halt if I told it to bake everything. to this end i had to split the model into groups which i then baked out seprately. This was fine, I added some overlapping areas to the bakes to preserve AO baking and the individual bakes worked well. Comping it all together was a bit of a pain though.
I got round this problem in some small way by rendering as PNG's with alpha and using the max composite node to combine all the baked maps together into 1 tile. It was a little slow to setup but it only had to be done once and it meant I didn't have to keep photoshop open in order to update my normal maps. I think im going to write a script to automate this process, it should be pretty easy to get all the RTT maps from a group of objects and create a comp map of all the different passes.
So how do you approach this sort of thing? Any sage advice for someone trying to streamline their workflow
Replies
this way you can render everything in 1 pass , its much faster and when you do testbakes you can imediately see what areas need fixing ( support geo , etc )
Then for the ao , you just need a self oclusion ao bake, exporting the non exploded lowpoly to xnormal simple ao generator does the trick and its super fast
Hope this helps !
If I have to explode your meshes for baking, set them to keyframes, so you can scrub back and forth on your timeline, and not have to dick about with multiple versions of meshes.
I apply materials to my HP mesh that represent materials it'll end up with. I bake these out, with diffuse, spec and gloss and use those as bases to add to, rather than using your masking material method and having to start from scratch with defining materials.
Johny: thanks man, i tried the render itterations trick with Turbosmooth but it still crashed out. im using a system with more ram now tho so ill give it a go.
That Self occlusion trick is damn clever! thanks for sharing
glynn: may i ask how you bake out the gloss values? i don't get a gloss element in RTT
I do this too. It's extremely helpful and doesn't take a whole lot of time at all. Making the cage is easier too.
you've seen this page? the eq ao method sounds like what you're after.
http://wiki.polycount.com/Ambient%20Occlusion%20Map?highlight=%28ambient%29|%28occlusion%29
hope it helps.
While this may be a good way to start, i feel like you'de do even better if you baked a RGB mask to go along with these base mats, as unless you always create perfect basemats and have no need to tweak, ever, you'll end up re-creating layer masks manually in the texturing processes. Which would be much slower than simply doing the RGB thing in the first place i would think.
frame 0 = original element positions
frame 1 = exploded element positions
you may need to select 'All' within the keyfilters to key the mesh. I think Autokey should know what to do regardless
the reason i like the layer mask approach myself is the quick tweaking. its more like offline rendering, change a value and press render
If you want to use a cage you have to set up the Projection modifiers individually though, because Pick makes the modifier an instance, which makes it grab all the targets at once (a total mess).
There's no gloss. Really, if all you need is a greyscale value, you could assign that to alpha/self-illum, if you're not using those.
Actually, yeah. I've always hand made my masks, which is a bit silly, thinking about it.
It never takes that long, but I guess every minute saved faffing about with something is a minute you can spend on something else.
its only 1 line in the RTT macroscript and no more waiting ages for a crappy cage.
if you do the following:
in the listender:
backup the script
go to line: 4785 which reads comment that out and save, then evaluate the script.
Now you can do the cage yourself.