• Announcements

    • Xiola Linden

      The New Community Platform   03/21/2017

      We are still working on making adjustments and changes to the new platform. Thanks  to everyone who has been sending in feedback and filing any bugs you've encountered! 
Sign in to follow this  
Followers 0
Syle Devin

Quality vs Land Impact.

23 posts in this topic

So I was curious as to what others oppinions were on quality of meshes vs land impact. What I wonder is if a couple more prims hurt if the quality increase is good enough. See I have this chair I have been working on.

chair.jpg

 

The mesh on the left is 4 prims while the mesh on the right is 1 prim. If I didn't say anythign could you tell that the one on the right had more of a rounded edge and other things? Is that type of quality increase matter if I can make the chair close to as good with 1 land impact instead of 4. 

 

Maybe I am just having trouble understanding how some people can get so much quality into such low prim objects. For example this couch looks like it has a lot of faces in the cushion yet it is somehow only 7 prims.

https://marketplace.secondlife.com/p/LISP-Mesh-Anna-Sofa-Texture-Change/4209156

 

I am trying to think if it might even be worth it to make a 1 prim version for any furniture that I might sell. Are there any tricks to squeezing as much mesh into a single land impact as possible? I know a few already but I can't think of to much else yet I see so much low prim furniture with so much detail.

 

What are your thoughts on this? Looking forward to reading the comments :)

Share this post


Link to post
Share on other sites

Ha yea I get ya. The thing is I've made mine more efficient but I don't know if I like how it looks. I'm trying to find out if people think a couple of prims, when it comes to three or 4 overall, really matter.

 

Share this post


Link to post
Share on other sites

From your pics, I honestly can't see a difference, admittedly I haven't got my reading glasses on but they both look fine to me.

Share this post


Link to post
Share on other sites

The difference is clear to me and will be even more evident in closer-up view. The secret of maintaining the quality of the higher detail mesh while getting a low LI is in careful design of the lower LOD meshes. Use the lower detail mesh, with the sharp edges, for the medium and/or low LOD slots, after deleting all those faces that will be hidden. Then make an even simpler mesh, by merging the cushions and the chair, for the low and/or lowest LOD - just a cube with the smaller cube subtracted (you can use that for the physics shape too). For objects this size, the lowest LOD has by far the most influence on the LI, and the low LOD the next important. So making these very simple will lower the LI without affecting the high detail view. You should be able to get this down to 1.0 LI without sacrificing any of the high LOD detail.

Share this post


Link to post
Share on other sites

I can see the difference clearly.

What's more I can guarantee you that your 4 LI chair CAN be made for 1 LI :D so no need to fret.   I have spent a lot (really a lot) of time working on just that. I got some tips from this thread ----   http://community.secondlife.com/t5/Mesh/Almost-Free-Land-Impact-How/td-p/1757119

 

The answers are all in there (at least some huge clues - my synopsis in the end will most likely be helpful), just read carefully :D.

 

You might want to make the cushion a separate "blender object" (not sure I have that terminology correct) and the rest of the chair another (each .5 LI) Just thinking about the animations. This may not be all that important; I don't do much of that these days. Good luck.  Watch EACH part of your upload costs. There are tricks for each area to get the LI down. Put them together? You have it made.

 

 Edit: While my experiment with the cubes is interesting and I have found it relavent, it really doesn't have anything to do with your chair build. Didn't want to confuse you on that.

 

Edit 2: I see while I was typing more info came in.  Drongle is no doubt absolutely correct with the different LOD models as the "best and correct" way to accomplish what you want. I will admit however that I have never yet made a LOD version and still have fantastic (finally) LIs with great LODs. So work on different methods and find out what works for you and TEST TEST TEST out on Aditi. You will be amazed with what you can accomplish.

Share this post


Link to post
Share on other sites

We tend to forget that sculpties are just one prim and make perfect sofa's since they are viewed at close range.

Mesh is perfect for larger builds that require a physical shape, doors and windows.

I am surprised that many creators import Mesh while many times a sculpt is a better option.

Share this post


Link to post
Share on other sites

Im working on it, I've gotten it down to two prims with the higher quality mesh. I know I can do better and get it down to one! Thanks for all the advice, it has been helpful. Setting up my own lods has helped a lot.

 

Personally I would never be be able to make something this good out of sculpts. That is just my own skills. Plus I find mesh so much easier to work with. If I can learn how to make mesh such as this into one prim then it would be just as good as a sculpt imo.

 

Edit: I got it down to one prim! I even was able to have the first two levels the same. I found it odd how much it seems the "lowest" level affects the land impact. Also I will do some testing about uploading the different objects as seperate meshes because I was noticing some weird things. For something the size of a chair that last "lowest" level is basically null and doesn't really show up anyway. Well that is what I got from it. I learned some things about making my own LODs which I had never done before. Thanks to those who helped :D

 

 

Share this post


Link to post
Share on other sites

As others have stated, you can certainly get that particuar chair down to 1 LI.  It's all in how well you manage the lower LOD's. 

That said, let's talk about your other questions, because the subject in general is an important one. Success in modeling for real-time environments is always about striking the best balance between quality and resource usage.

With that in mind, let's pretend for a moment that there was not actually a way to lower that chair's LI, and that it really had to be 4.  If that were the case, then the question you'd have to ask yourself would be do the rounded corners make it look 4 times better, in order to justify 4 times the overhead?  If it were me, I'd say no, they don't, especially when you consider that people are likely to have a lot more than just one instnace of any given chair in any given scene.   While it does look a little better rounded, it doesn't look 4 times better.

That's the way I generally approach the balancing act.  Really, it's just like balancing any other budget.  If you're getting a good return on your investment, then it's worth doing, and if you're not, then it's not.

Here's the logical thought process I use, when deciding whether or not to make something look better, at the expense of higher resource usage.  If I can make it look a lot better, while only costing a little more, then it's generally worth doing.  Conversely, if it's only going to look a little better, while costing a lot more, then it's generally NOT worth doing.  If the factors are equal, meaning it's either going to look a lot better while costing a lot more, or look a little better while costing a little more, then it MIGHT be worth it, depending on what else is going on in the game.

If it helps, here's the same set of factors and conclusions, diagrammed:

 

quality vs cost 1.jpg

 


Of course, the reverse is also true, when deciding whether to reduce resource costs, at the expense of quality.  If I can make it cost a lot a less, by only looking a little worse, then it's worth it.  Conversely, if it's only going to cost a little less, but will look a whole lot worse, then it's not worth it.  And again, if the factors are equal, then it's a maybe.  Here's the same kind of diagram, for this situation:

quality vs cost 2.jpg

 

As passionate artists, we have a deeply personal investment in how good our artwork looks.  And as wary technicians, we have an equally strong concern for how efficient our assets are.  As human beings, it's easy to become overly emotional about either one, to indulge the one, while drowning out the other.  Whether we're aware of it or not, we tend to tie our sense of identity to these things.

To ensure our decisions are pure, it's important to rely on a set of guiding principles/  The solid logic in the two diagrams never fails.  By placing each decision into one of the squares, to see how it measures up, we ensure the best, most balanced result, every time.  If it's in a white square, it's a good decision.  If it's in a black square, it's a bad decision.  If it's in a gray square, then you know you have to look at other factors*, and use the same sort of process on those, as well.

Make all your decisions in accordance with the process, and you'll never go wrong.

 

 

 

 

 

 

*Here are a couple of quick examples of incorporating other factors, when faced with gray-square decisions.

A few years ago, I was involved with creating a virtual world for users in an a developing nation.  The average computer in that country is about ten to fifteen years behind what we're all used to here in the first world.  Therefore, we had to do everything we possibly could to ensure good performance.  It was a constant challenge to figure out how to use as few resources as possible, at every turn.

The client was an advertising company, from the country in question.  They had no idea how the technology actually worked; they just wanted a virtual world for their country.  They supplied us with lots of 2D concept art, which their graphic artists had come up with.  Needless to say, most of their designs were not actually possible in 3D, given the technical restraints.  So, we had to find ways to do things that looked passably similar, while still being usable on a circa 1995 computer, operating at dial-up-like connection speeds.  It proved to be one of the most difficult of all projects in my career.

Needless to say, on this project, the "costs a lot less" category was king.  Anything from that first diagram was totally out, as there was no room for "costs more", ever.  The second diagram was the only thinking that could apply. 

We would take the concept art, build 3D models of what we saw in it, and then creatively study the results, to figure out how we could change it to make it viable.  Any time something fell into either of the "costs a lot less" squares, that was the way we had to go, even if it also had "looks a lot worse" attached to it.  We strove for the white square as much as possible, of course, but in most cases, the upper right gray square was the one.  On rare occasions, we could afford the lower left, but not usually.  Anything we could do to lower resource costs, we simply had to do, and that was that.  As long as things looked plausible, even if not good, they were good enough.

All things considered, the project turned out pretty well.  It actually looks pretty decent, for what it is, and it's still in use today.

 

For an opposite example, I was once involved with creating a virtual dating service.  The idea was after you connect with someone on a dating site, but before you meet them in person, you first go on a virtual date.  Since the environments would be small, and since there would only ever be two characters at a time, we were able to pull out all the stops, in terms of visual quality.  Almost everything was able to fall into the lower left square in the first diagram.  Things could cost a lot more in order to look a lot better, and that was just fine. 

The character models were equivalent to DAZ Victoria/Michael quality, to give you an idea.   That's typically unheard of in games, but when you only need two of them, it's doable.

Unfortunately, the project was derailed before it got off the ground, so we never got to finish it.  A less than scrupulous programmer ended up spooking the investor into cancelling the whole thing.  It's a shame.  It would have been pretty cool.

 

In SL, gray-square decisions aren't so simple.  It's a chaotic environment, in which little is truly under your direct control.  The most important question, really, is how does the item's appearance impact the user experience.

In cases where other things are going on that more profoundly affect the user experience than the looks of that one item, then it's best to keep the cost as low as possible, even if that means sacrificing quite a bit of visual quality in the item.  That's especially true if it's something people are likely to rez multiple copies of, like a chair.  If I'm having a bunch of people over to my spiffy new castle in the sky, to enjoy an evening of socializing, or a trivia game, or what have you, does it really detract from that, if my chairs have square corners instead of round ones?  I'm sure you'd agree, the answer is very likely no, it doesn't make any difference at all.

On the other hand, if it's something whose looks are what directly make or break the user experience, like an exquisite sculpture in the middle of an otherwise empty room, then it might be worth it to jack up the cost, in order to really put the spit & polish on it.  In that scenario, clearly, the visual quality of the item IS the experience.

No matter what the scenario, though, at the end or the day, it's always about doing more with less, as you already know.  In most cases, that just comes down to application of a little knowledge plus a lot of common sense.

1 person likes this

Share this post


Link to post
Share on other sites

"I found it odd how much it seems the "lowest" level affects the land impact."

If you want precise details, and don't hate mathematics, you could look at this old thread. In the extreme case, you could use just a single cube with no bottom and a texture to suggest the shape, at the lowest LOD. As you say, you rarely see it. Don't forget though that some people will be using low graphics settings and might see the lowest LOD nearer than you do. It's sometimes worth testing this by looking at your mesh using low graphics quality.

1 person likes this

Share this post


Link to post
Share on other sites

Hmm well I already turned the lowest level into a cube with no bottom but I did not think about others who might have lower graphics than I do. I'll try and see how it is. One other thing though, does anyone know if it is possible to have different textures for each LOD distance? Such as where the original texture map is on the highest quality but on the lowest quality that is only a cube I might have a modified texture map so it still looks like the chair and not a cube of fabric.

Share this post


Link to post
Share on other sites

Yes. It is possible. The requirement is that you need to have the same set of materials used in each LOD*. You can achieve that by having a small hidden triangle in the high LOD mesh to carry the material for the lowest-LOD texture, and also one(or more) in the lowest LOD for unused high LOD texture(s).

*If you study the wiki, you may notice that the stated requirement is that the lower LODs can have a subset of the high LOD meterials, implying that the extra triangles are not needed in the lowest LOD mesh. However, there are some bugs that cause the wrong textures to be applied after LOD switches to higher LODs in some circumstances. To avoid these you need exactly the same set of textures. As far as I am aware, these bugs haven't been fixed, but after the jira accessibility changes, it's difficult to know without constant re-testing.

Share this post


Link to post
Share on other sites

You can incorporate different textures per LOD mesh, BUT this will require using a separate material per new texture to do so. Keep in mind that a mesh can hold a maximum of EIGHT materials in total - this includes all the LOD meshes.
In theory, one dedicated texture for your maximum detail mesh and subsequent lower LODs, and a separate dedicated texture for the lowest LOD would be sufficient. It's always a good policy to aim for the best possible efficiency, so a lower resolution texture for the lowest LOD mesh would be a good idea, especially since it would only be visible at a longer viewing range (and the pixellation would not be easily discernable).

Each LOD mesh will need to have each material in usage, even if not actually intended for that particular LOD. A good method is to place each of the unused materials per LOD on tiny triangles and hide them from view inside the object. By doing so, each LOD mesh can, if necessary, have its own dedicated texture.

However - the added materials and requirement for hidden triangles will possibly increase your Land Impact calculations slightly. Not by much (if at all), but keep this in mind. The gain in visual quality, though, should be more than worth it.

:matte-motes-smile:

ETA: Argh, beaten by Drongle AGAIN! :matte-motes-wink:

Share this post


Link to post
Share on other sites

Loved Chosen Fews graphics. My old logics professor would have smiled.

For me, and this subject is VERY important to me -- the primary thing to remember is that no one is going to see the hinges on your door from way across the sim. I spend a lot of time camming out from a distance (camera restraints is where you get that ability - in preferences but I imagine folks here know that) to see how things look.

It is very true that many (maybe a majority since I haven't seen so many of those "how to make your sculpts look better notecards") may have their LOD viewing set at the default which for some viewers is 3. Firestorm now has a very easy way to up that to 4 -- right in the Quick preferences, but lots of folks most likely don't know what LOD means and why they should care.

So whenever possible (and it often is ) I set the LODs for the two highest levels the same. On some items (again visually testing) you can have the LOD at 0 and the object will still hold its shape from beyond sim distance. And honestly, who in the real world could possibly see that well? We know a car is a car not because of the details, but because we perceive it as a car -- often a moving one.

So the hard decisions really come with larger structures that are more costly in the LOD mid and low range. It might be very important to you for your arches to stay crisp. What I have noticed in the RETAIL world though, is that the top mesh designers opt for a middleground. So a house WILL fall apart a bit from way across the sim. But most of the time we see it from a closer distance. I have a tendency when faced with an actually NEED to make that quality vs LI decision (all the tricks to get the LI down have been exhausted) to go for the quality. It is often only 10 percent of the LI cost will get you a much nicer build.


I wrote an article recently on my design blog about mesh and land impact. It might be of interest. Don't be put off by the scantily clad photo -- it really IS an article all about mesh *wink*. Gotta keep the fashion feeds happy. They are not big on techie prose.

While LI will most likely not be important in the art world, if we are talking products -- I think the future belongs to the designers that can make smart mesh and texture well (that let's me out as I am not a great texturer :D). When you can live in superb style on a 512 for free by buying beatiful low impact furnishings the need for all those prims kind of fades away.

 

http://chicatphilsplace.blogspot.com/2012/12/into-future.html

 

Share this post


Link to post
Share on other sites


Maeve Balfour wrote:

 

However - the added materials and requirement for hidden triangles will possibly increase your Land Impact calculations slightly. Not by much (if at all), but keep this in mind.

There are a few more things to take in mind when you use different textures for every LoD.

First is your landimpact, which shouldn't be the first thing to optimise, but often is. If you use a seperate material for every LoD, you will have to add an extra triangle for every unused material in the other LoDs. So if your lowest LoD is a bottomless box which has 20 vertices (if the edges are sharp) and 10 triangles, you'll have to add three triangles with 3 vertices each. This means an extra 9 vertices and 3 faces, quite a lot more. Depending on the size of the object and the numbers of it in a linkset,  this can result in a substantially higher landimpact.

Second is texture use. Although it might not be obvious at first thought, adding a low resolution texture to your lower LoDs will not lower your memory use, it will raise it slightly. Because of mipmapping, SL won't download the full texture if the object is far away, but a sampled/scaled down version. Adding an extra texture will...well add an extra texture to be processed. It doesn't really replace the others.

Third is the ability to modify colour on an object. If you have for example a car which has a shiny material for bumpers, a flat material for paint and a translucent one for the windows on the highest LoD and a small texture for the lower LoDs, changing the colour of the paint won't affect the color on the lower LoD material. Changing the colour on the lower LoD material will change it completely. So your chrome bumpers and glass will change colour with the paint. In such a case it's probably better to share materials between LoD models.

 

 

2 people like this

Share this post


Link to post
Share on other sites

Chic Aeon wrote:

So a house WILL fall apart a bit from way across the sim.

It doesn't have to and it shouldn't. The trick is in making good use of linksets. Maeve has tested a lot with this. I myself have done some. My first "rule" is to break your build into exterior, interior and windows.

The exterior should be very LoD resistant to keep the shape over a distance. A big build will never fall to pieces with just the highest LoD modelled. The rest can be set to 0. (This is with all the default LL settings from low to ultra , a very long draw distance in combination with a low object detail won't work).

The interior can fall apart from a distance any bigger than the biggest indoors distance it will be seen from. This means you can break it into pieces. The entire interior as one mesh results in a LoD switch distance far bigger than needed.

The windows are more or less flat, so you can use imposters at LoD med, low and lowest without it being obvious or even visible at all.

The second thing to do is cutting up/combining pieces. Currently I'm working on a building about 50x30x15 meters. The plan has an I-beam shape. I tried some different setups for the exterior:

- Making the entire exterior out of one piece

- Splitting it in three with the vertical part of the "I" as one piece and left and right each as one piece

- Splitting it in three with the vertical part as one and top and bottom both as one

- Finally I tried it as 5 seperate pieces

The first option results in very resistant LoD behaviour, more than you need. It probably won't fall apart at more than 1024 meters. This is too much for the default settings and results in an unneccecary high LI.

The second option kept the shape of the outer pieces at any default LL setting, but the center broke. Rather than combining the pieces again, I added a medium LoD model.

In the third option all three pieces broke from a big distance. Given the bigger complexity of the outer models, adding a model for medium would have resulted in a high LI. I didn't try this.

The same is the case for the last option, but of course the models broke from a shorter distance.

All in all, if I had fiddled with the medium LoD models to let them keep their shape, the different setups would have looked the same, but the LI would have a factor 3 between them at least. With option 2, I managed to keep the LI for the entire exterior at 15. Uploaded as one single object it would have been 20.

Building_Ext.JPG

I am now fiddling with the windows, looks like clusters of 3x3 for the wings and 2x2 for the ends will result in the lowest LI.

Also notice the geometry for the entire exterior is about the same as half a sculpt for the tris and about three fourths sculpts for the verts. Just to point out how horrible sculpts are.

2 people like this

Share this post


Link to post
Share on other sites

The only problem with multiple parts for an exterior is that, from angled views, the parts swtich LOD at different camera locations, so that you can see mixed-LOD views. This can be unpleasant unless due care is taken with the medium LODs to mitigate the effects.

Share this post


Link to post
Share on other sites

:D. I actually knew it didn't have to fall apart *wink*. I am a small home gal so I doubt I will need to be tricky, but I do understand the separation of parts as I used that even on my tiny casita and saved a TON of LI points.

I think experimentation and testing is definitely the key. My little house (and it is little ) holds together across a sim and that is good enough for me.

I also think though, philosophically, that it can be easy to get very wrapped up in the quest for perfection. I had an occassion today where I had  to decide whether to upload at 1 or 2 LI  for a fairly complex piece of furniture. In the end I opted for 1.5 which if someone wants can easily be 1 by making it just a tiny bit smaller, or can be much larger and retain the 2. It holds it shape half way across the sim and honestly, it is supposed to be INSIDE the house - LOL. So that's good enough for me. The first two LODs are full.

If you are building a castle, that will be seen "across the land" (mainland that is) then I can see the importance of a really good low LOD, no aurgument here. Many times, though, it really doesn't make much difference.

Share this post


Link to post
Share on other sites

This is a life lesson hidden in a second life lesson.

Don't get so caught up in the trees that you become lost in the forest.

Share this post


Link to post
Share on other sites


Drongle McMahon wrote:

The only problem with multiple parts for an exterior is that, from angled views, the parts swtich LOD at different camera locations, so that you can see mixed-LOD views. This can be unpleasant unless due care is taken with the medium LODs to mitigate the effects.

There is no switching at all, that's the point. The switching happens out of range of any default settings.

You don't want to match the switching, you want to eliminate it altogether, either for real or just how it appears.

Share this post


Link to post
Share on other sites

"The switching happens out of range of any default settings."

I see what you are aiming at. I was considering smaller components. But, of course, the download weight is designed to take account of this. If your pieces are large enough to never switch, then they are large enough that even the medium LOD has litlle effect on the LI. Most of the default settings steps use RenderVolumeLODFactor = 1.125 and RenderFarClip = 128 (except for Ultra and Low). That means you will be able to see the first switch within the draw distance unless the mesh has a bb diagonal about 55m or greater. That's just about the size where triangles are as expensive at medium as at high LOD. So if you want to have much smaller pieces, so that you can save LI with a good medium LOD, then you do need to be careful about the medium LOD of those smaller pieces. I think that's consistent with your building, if your parts were the uprights and the crossbar of the 'H'. The diagonals of the uprights would not be far off 55m. The crossbar is smaller andI think you said it required attention to the medium LOD, as I was suggesting.

Of course, these numbers are purely theoretical. The fact that you can see a switch if you look for it doesn't mean you will in normal circumstances, when you are busy flying towards or away from the object and looking at nearer things. So you probably have a bit more leeway before the switch effect becomes unpleasant.

 

1 person likes this

Share this post


Link to post
Share on other sites

Wish all objects could be like that. Size seems to affect that such as my chair, being small, switches at quiet a distance but still not far enough that you wouldn't notice. Even with odd switching for smaller objects I feel that it can be done nicely if everything is planned out. I personally wish, atleast to even try it out, that sl acted like a lot of games that just don't rezz things from certain distances because you can't see. Meaning there is no low poly it means there is no object. So if someone set their regular view distance to 80 meters, everything in that 80 would be perfect but you wouldn't even see anything outside of the 80. You would have to set your view lower most likely but the game would look better possibly. Just throwing out an idea...

Share this post


Link to post
Share on other sites

I don't think that is true at all, I have played plenty of games where there are multple LOD's and you can see them switch as you move towards things. This is especially visable in large open game worlds like Skyrim and GTA4 for example.

Share this post


Link to post
Share on other sites


Syle Devin wrote:

So if someone set their regular view distance to 80 meters, everything in that 80 would be perfect but you wouldn't even see anything outside of the 80.

If LL did that, you'd have a terrible SL. Imagine all 1 meter big objects at their highest LoD from 80 meters. The amount of triangles to render would be immense. It's up to the builder to make sensible LoD models, so everything looks good from any distance at any setting. For some odd settings it's simply not possible, but I always try to get good looks at all the default ones.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  
Followers 0