Question regarding your asset building process
polycounter lvl 7
Hey guys, let me just start by saying I love Polycount. I've been enamored with the forums ever since stumbling across them a few months ago. This is a really awesome place.
I've been working with UDK for a few months and just recently getting into asset creation (using 3ds Max 2012). I've watched tutorials on 3dmotive, Eat3d and hundreds scattered around the internet [especially Thiago Klafke and Philip K - love you guys!], but I'm a little confused when it comes to the 'proper' asset building process. Obviously there is no 100% correct way to do things, but there are best practices and that's what I'm curious about.
Anyway, my question... when you're creating an asset, do you first create the texture, then build the geometry around the image or build the geometry first and create a texture to fit that geometry? Or does it just depend on the situation?
I've seen a number of tutorials that suggest it both ways, but when trying each method for myself, I've also run into complications, so it's not clear to me which way. I'd like to hear what you all think is the best method.
Thanks
I've been working with UDK for a few months and just recently getting into asset creation (using 3ds Max 2012). I've watched tutorials on 3dmotive, Eat3d and hundreds scattered around the internet [especially Thiago Klafke and Philip K - love you guys!], but I'm a little confused when it comes to the 'proper' asset building process. Obviously there is no 100% correct way to do things, but there are best practices and that's what I'm curious about.
Anyway, my question... when you're creating an asset, do you first create the texture, then build the geometry around the image or build the geometry first and create a texture to fit that geometry? Or does it just depend on the situation?
I've seen a number of tutorials that suggest it both ways, but when trying each method for myself, I've also run into complications, so it's not clear to me which way. I'd like to hear what you all think is the best method.
Thanks
Replies
However I am super new to this, so hopefully my reply will spur someone with better knowledge to give you a suitable answer
placeholder/blockout
highpoly
lowpoly
texture
I've seen people do a texture first style workflow. However it doesn't always give the best result in my opinion since you can't get as precise normalmapping on the models.
There is indeed a method of creating one "all inclusive" texture map and then adapting multiple, separate meshes to use it. The idea is to reduce the number of texture maps needed to render your level properly which greatly increases performance. However, this method doesn't really work for complex models (like characters) because their texture needs too many unique features and their geometry is too complex to get proper alignment. This method is mostly used in situations where optimization is drastically more important than quality. I've mapped small little "clutter" objects this way for example.
So basically, gauge the importance of the model you're working on. Would a player stop and look at it or is it just there to fill the scene? Also, how comfortable are you with the UV mapping process? Knowing that like the back of your hand really helps with situations like these.
Check out Tor Fricks sci-fi environment. He made a single small texture first and built a whole environment with it.
For me it depends on the asset. If it's a unique piece, I bake from a high poly to low and paint a texture afterwards but if it's part of a modular set I'm more likely to develop the texture and the models at the same time and spend a little more geo to make up for using CB instead of baking from a high poly.
Having fewer textures in memory doesn't always increase performance. It just gives you more free memory. Also this style of construction can create a lot of UV splits which will use more memory. This is particularly troublesome on platforms like the PS3 where the vertex memory and the texture memory are sometimes two different things. So in that case you can't trade one for the other.
However for low end platforms like a phone it may increase performance. It depends on how the renderer handles it.
Getting my head around UV mapping has been one of the 'complications' I've run into. For example, I think I'm pretty decent at creating flat textures in photoshop and applying them directly to a flat planes or rectangle geometry (using ndo2 to fake depth), because there's no stretching or unwrapping involved. However, when I try to create a more complex static mesh, it has to be unwrapped and so I 'flatten mapping' and end up with my object split into 30+ pieces, all laid out onto the UVW render in a way that takes up very little space, but is very difficult to create textures for. If I've already created a texture, now I have to cut out the individual pieces and try to line them up. If I haven't created a texture, then I'm trying to create one that lines up correctly across the flattened (and scattered) pieces. Does that make sense?
I feel like I'm doing something wrong, but only because it seems harder than it should be. All of the tutorials I've seen, always seem to cover either the unwrap process or the texture creation process, but never both, within the same tutorial.
I was so excited to see this included in the March release of UDK - I've been following that thread for a while. I'm a sucker for anything Sci-Fi. Getting to 'play around' and see how he did that was awesome.
Don't use 'flatten mapping'. Learn to unwrap properly. It'll make texturing much easier.