Looking for Information and Advice Regarding Res LODs

[Rabid Squirrel 149]

Hi there,

 

I just recently started modelling and got my first Model into Arma 3 a little while ago. However, I am still a little bit hazy on how to put together Resolution LODs and I have a few questions/concerns:

1. I have found out a few things about resolution LODs, one being that each LOD should be half the size of the one bigger than it in terms of 'verts' (I believe). The biggest thing is: Do I have to model each individual Res LOD, or is there a shortcut that people use that I haven't happened upon yet?
2. I imagine I'm going to have to match the UV sets of the lower res LODs to match that of the higher poly LODs, does that mean I have to create individual texture sets for each LOD, or just have them use the normal texture sets.
3. Is there a benchmark for how many LODs I should have, and/or what is the performance impact of using less/more Res LODs?
4. For the guys that make tons of models, do you guys have a surefire way to quickly output the additional Res LODs, and what are they if you do?

The Model is an Assault Rifle, and it is likely that you are going to have a bunch of them on your screen at the same time, which I imagine will affect performance.

 

Thanks in Advance,

Squirrel.

 

[da12thMonkey 1943]

Res LODs can be a bitch. How to make them kind of depends on what software you're using.
If it's only Object builder, your options are kind of limited to either remodelling and re-UVing, or merging verts together so that the faces between them collapse and the UV space sews up too.
 
Other 3D programs provide you with the option of selecting entire edges, deleting those and then deleting the verts that used to form those edges and letting the mesh relax back between the new endpoints of the edge. So long as you steer clear of deleting UV border edges the UVs will relax inside the boundaries and you shouldn't have to do much to change the UV maps except occasionally to sew up holes.
 
As I say, it depends on what your 3D program is like. Personally I only know Maya, and have some insight in to 3DSMax from listening to a bunch of people who use it while we discuss modelling topics, so I know it can do basically the same process. But I'm pretty sure blender and other software are capable of the same methods of manually removing edges and then the extraneous verts without having to remodel the entire bloody thing and make new UVs that align with the original ones.

 

To show an example of this I'll show you how I reduced certain parts on a model I have.
In the base there is this dense topology in the pilot view and top could of visual LODs where I have added a bolthead that is sunk in to a hole. 1
 
I've UVed this in a way where the UV island for the face of the bolthead are placed inside the corresponding holes in the base so that if I remove the bolthead geometry and fill in the hole, it will still look like it is there in the texture. 2 3
 
I feel this is a pretty good area to reduce the face count while maintaining the appearance of the original model from a distance. So I go about removing any more faces attached to that area that wont be useful when I close it up. 4
 
And that leaves me with the above ring of verts on the old UV border edge (the edges that are highlighted as slightly thicker, are the border edges) that I can start to merge together to reduce the number of points. 5
 
But you'll notice that the border edge verts don't move in the UV space when they are merged. Therefore I have to sew the UVs for those points together along the edge I just closed together, so that it matches the model topology again and doesn't distort the texture. 6

Notice that the edge is no longer highlighted with a thicker line, now that it's sewn together and no longer forms a border edge.
 
I can then repeat that on the other side and I'm just left with a UV island that has no internal UV borders. That means I can just delete any of these extra edges wholesale and then remove the verts that were attached to them to form nice tri/quads instead of nGons. 7
You'll see at the bottom though there are two verts that didn't delete because there is a connecting edge on the bottom side of the mesh. So then my next task would be to remove the unnecessary edges on that side of the model or reconnect verts that cannot be deleted because they form an outer border edge of the UV island in oder to get rid of nGons.
 
As I said, you want to avoid deleting the outer UV border edges unless you're deleting the whole part of the mesh that connects to the other side of that border (like the bolts at the start of my example). If the other side of the border exists, then the UV map will try to bridge the gaps across the UV space between different UV islands 8 9

 

My personal recommendation is to do that kind of process for the first LOD or two to make sure you have maximum control over how the mesh is decimated, and the silhouette remains intact. But after that, most mainstream 3D apps have a "reduce" modifier of some sort that will delete edges, merge verts and generally reduce the polycount of an object for you by a set amount (usually offers to do it in terms of percentage).

However the results of doing this can be a bit messy with complex models and merge parts together that you don't want to merge together if you're going to animate the model. So you need to be careful. (Same model from the start but just applying a 50% reduce modifier to the whole thing 10 face count is reduced to an appropriate degree but there is no real logic to the meshflow anymore)

You can do this from the get-go if you're none too concerned about appearance after the top LODs though and just want to do it in the most fast and lazy way.

 

But, there is also a great thing called Simplygon that can yield better automated results than the reduce tools in most 3D apps https://www.simplygon.com/and I know a bunch of people who use this.

[Rabid Squirrel 149]

Thank You for the Information

 

I'll play around with the model a bit and check out 'Simplygon'.

[lodu 145]

For my part, in addition to da12thMonkey writes, I am very careful deployment of uv in the first lod, so that optimization / decimation in the second lod, occurs naturally and I did not really need to intervene (I speak especially of organic assets, since it is the essence of my creation: characters, plants, rocks, etc.)
With my personal experience, I would say there are as many methods of decimation, that objects and often you do some work with a dedicated tool and you complete, end "by hand".

I work on the blender and decimation tool is very useful, but I also prepare my model, to help the tool to "find the right path."
For example, on a tree trunk, I'll prune small branches at the end of large branches, so that the tool does not take into account and focuses only on the bulk of the object: the calculation tool is global, having removed these small branches, which anyway does not see the current lod because they are hidden by the foliage, provides good and quick decimation.

So I spend time thinking about the location of my uvs on the architecture of the model to simplify as much as possible: we must constantly think in the depth of the object, on its first three lods, closely nested, particularly because the figure should remain on about a lod identical to another while decimating each time by almost half.

Depending on the object, I prefer other tools, such as "MeshLab" extremely effective for achieving lods rocks (and certain other industrial forms).

To stay in the vegetation, when I make a branch with foliage above, I make sure to also prepare the other two lods inside, which I attibuer a named temporary selection, which allows me to position a duplicate branches number of times, in many different axes to shape the mass of foliage that makes up the tree, then select the party named A2, for example, that I copy / paste in the lod 2 and part I A3 copy / paste in the lod 3: graphical continuation of the tree silhouette is provided for these three lods and the fourth and last lod has some polyplanes that simulate the tree in the distance.

So too, there is a continuing reflection in depth (ie: different lods), which creates no hair pulling, which can probably be applied to other types of objects.

I think it is important not to see an asset by limiting his first lod, without finding out what happens next, once this first lod is over, but, instead, see addon in its entirety.

[x3kj 1247]

also something to keep in mind for lower lod:  arma produces alot of z-fighting issues for distant objects. Faces that are behind each other will cause these issues. And we are not just talking about some mm distance here, i've had stuff z-fight with face distances > 12cm.

You can see it for example on vanilla infantry. Their bodyarmor z-fights with the underlying clothes at larger distances.  So keep your lower LOD very clean and solid and without much overlaying parts to reduce this.