Jump to content

What is a good amount of triangles/vertices/faces for a mesh body?


TaleTitan
 Share

You are about to reply to a thread that has been inactive for 110 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts

30 minutes ago, Gabriele Graves said:

I don't understand what you mean, perhaps you can clarify?

You said: "It is easy to decry that this is because of the insane amount of triangles going into mesh avatars and that this shouldn't be happening.  I can accept that but I am also told that this can all be optimized without noticeable quality loss too so I am not sure what to believe there."

I would say: The insane amount of triangles shouldn't be happening because in the vast majority of cases it can be optimized/reduced without noticeable quality loss.

Unless I'm reading it wrong and what you're saying is that "if it's that bad or easy to fix, it probably wouldn't happen as much. There might be a reason for it."

Edited by Wulfie Reanimator
Link to comment
Share on other sites

3 minutes ago, Wulfie Reanimator said:

You said: "It is easy to decry that this is because of the insane amount of triangles going into mesh avatars and that this shouldn't be happening.  I can accept that but I am also told that this can all be optimized without noticeable quality loss too so I am not sure what to believe there."

I would say: The insane amount of triangles shouldn't be happening because in the vast majority of cases it can be optimized/reduced without noticeable quality loss.

Right, perhaps I just worded that badly.  The part about "not sure what to believe" is because we cannot flick a switch and see those potential optimized savings whilst seeing that the avatar hasn't discernibly changed and that part is hard to believe for me.  Showing me examples doesn't help because my brain wants to reject those as contrived and so it is a hard sell.

  • Thanks 1
Link to comment
Share on other sites

Maybe it might help to look at popular bodies in wire-frame mode. I stole some pictures from other people to try to make a point.

wireframe1.thumb.png.578f035dddca4d26f4000615c8ea8568.png

Stolen from: https://avastar.machinimatrix.org/279/help/toolshelf/rig-converter/convert/

wireframe2.thumb.png.453424171272bbbea1f008111894303c.png

Stolen from: http://blog.nalates.net/2015/04/15/second-life-mesh-body-revolution/

(Also an actual article about mesh bodies that might be of some value. I'd consider the Slink body to be optimized compared to others.)

wireframe3.png.7b352894d79809fad46c5064541f580d.png

Stolen from: https://catnapkitty.wordpress.com/2015/12/05/how-to-not-buy-bad-stuff-in-sl/

Consider if all those polys are really necessary, even when you are looking at the body somewhat close. Would any noticeable detail be lost if the polygons were reduced in certain areas? I'm sure certain areas need more polygons (curves, bendy points) than others. It's a balancing act: You don't want your body to look like the crappy SL default avatar, but you don't want your mesh to be so dense that it contributes to lag problem. 

  • Like 2
  • Thanks 1
Link to comment
Share on other sites

10 hours ago, Pirschjaeger Fassbinder said:

(Also an actual article about mesh bodies that might be of some value. I'd consider the Slink body to be optimized compared to others.)

I don't have the regular Slink body but one of my alts has the Slink Hourglass

The body itself without hands and feet are 53,660 tris. That is with tattoo and two clothes layers though. So each layer has 13,415 tris. That's about what I expected to be a well optimised body.

The hands and feet are a bit harder to evaluate though. Each hand actually has more tris than the body, 18,645 and 19,011 respectively and I'm not sure why. The mesh doesn't seem to be that dense in wireframe mode:

bilde.thumb.png.47833665122dd8e791f1666122c2af34.png

There are some obviously superfluous tris - 328 tris just for the anchor prim (or mesh) and 1,740 for the five leftover wrists from the days you used Slink hands with your system body (then again, maybe some people still do) and those three different nails with about 1,500 tris each. It's still the main part of the hand that accounts for most of the tris though.

But with that being said, the high tri count for the hands shouldn't have mattered since the LoD system is supposed to take care of it. Unfortunately it doesn't because of those pesky fitmesh LoD bugs. That's not Siddean Munro's fault, those hands date back to long before anybody were aware of the bugs.

It's a similar story with the feet. They're actually a whopping 138,716 tris for the pair. But that includes five separte meshes for different feet angles and only one is render at any given time. Ideally you'd have separate pairs to wear for each angle of course but it shouldn't matter much for the actual render cost - although it does increase the streaming cost quite a lot. And just like the hands, it's the LoD bugs that ar the problem, not the tri count in itself.

 

10 hours ago, Pirschjaeger Fassbinder said:

Consider if all those polys are really necessary, even when you are looking at the body somewhat close. Would any noticeable detail be lost if the polygons were reduced in certain areas? I'm sure certain areas need more polygons (curves, bendy points) than others.

Well, compare Slink to the others. Is the visual quality significantly lower than the others? I'd say no.

 

9 hours ago, Janet Voxel said:

Soooo, which one of y’all are making this low poly body?

Slink seems to be very, very close. With BOM eliminating the need for tattoo and clothes layers and with the render LoD bugs fixed, it would have been excellent.

Edited by ChinRey
  • Like 1
  • Thanks 1
Link to comment
Share on other sites

It's just too bad that the Slink bodies aren't very appealing to me aesthetically.  The HG is just the wrong proportions for what I want.  The Physique is closer and when I was choosing my mesh body I trialled demos from both Slink and Maitreya.  Initially I liked Slink because it seemed to be very close to the system body and did fit many of the mesh clothes I had already bought for that.  I even wore both bodies at once (without alpha) to see how close they were and found that the Slink body was a very close shape to the system body albeit much better quality.  However when I trialled Maitreya, the aesthetic difference was night and day and there was no competition in terms of what I was going to choose even though I knew my system mesh clothes were toast.

So that coupled with the fact they got rid of their alpha cuts makes them unsuitable for my uses.

I think this is a big part of the problem, there aren't enough choices available that are optimised well which will appeal to a broad range of people for the aesthetics.
Strangely enough though I used to wear their hands and feet like many other people long before I went to a full mesh body and even longer before I transitioned to a mesh head.

  • Like 1
Link to comment
Share on other sites

5 hours ago, Gabriele Graves said:

It's just too bad that the Slink bodies aren't very appealing to me aesthetically.  The HG is just the wrong proportions for what I want.

Yes, the HG body is a special one, not what most people are looking for. It's the Physique that is Slink's mass market product.

 

5 hours ago, Gabriele Graves said:

However when I trialled Maitreya, the aesthetic difference was night and day and there was no competition in terms of what I was going to choose even though I knew my system mesh clothes were toast.

The Maitreya body seems to have about twice as many tris as it needs. That's not really good enough of course but it's still far better than most other mesh bodies and indeed most SL meshes of any kind. And to be fair to Maitreya, we have to remember that it's an old body, released at a time when very few users and content cretors and no Lindens had any idea how devastating high poly meshes are for performance.

Anyway, I think I've found a final answer to what a good amount of tris for a mesh body would be. Here are the figures for Ruth 2:

  • Body: 11,410
  • Hands: 5,908
  • Feet: 6,896
  • Total without head: 24,214

With good LoD models and without the LoD bugs that should bring us well below 10,000 for normal view. Ruth 2 has a few annoying flaws, especially in the UV mapping and rigging, but those have nothing to do with the tri and vertice counts.

As for the head, take a look at this gallery of rouges. These are all system heads, how many more tris do they need? Maybe a few but not many.

mugshots.thumb.jpg.ce2c63a2580d0287133e9a4f2792da93.jpg

(Personal note: these are of course all pictures of some of my own alts, mostly shapes I made myself, spanning a time period of seven years. I always aim for max. variety and it's a little bit annoying to see how they all tend to gravitate towards a certain style. I suppose that's inevitable.)

Edited by ChinRey
  • Thanks 1
Link to comment
Share on other sites

3 hours ago, ChinRey said:

Yes, the HG body is a special one, not what most people are looking for. It's the Physique that is Slink's mass market product.

 

The Maitreya body seems to have about twice as many tris as it needs. That's not really good enough of course but it's still far better than most other mesh bodies and indeed most SL meshes of any kind. And to be fair to Maitreya, we have to remember that it's an old body, released at a time when very few users and content cretors and no Lindens had any idea how devastating high poly meshes are for performance.

Anyway, I think I've found a final answer to what a good amount of tris for a mesh body would be. Here are the figures for Ruth 2:

  • Body: 11,410
  • Hands: 5,908
  • Feet: 6,896
  • Total without head: 24,214

With good LoD models and without the LoD bugs that should bring us well below 10,000 for normal view. Ruth 2 has a few annoying flaws, especially in the UV mapping and rigging, but those have nothing to do with the tri and vertice counts.

As for the head, take a look at this gallery of rouges. These are all system heads, how many more tris do they need? Maybe a few but not many.

mugshots.thumb.jpg.ce2c63a2580d0287133e9a4f2792da93.jpg

(Personal note: these are of course all pictures of some of my own alts, mostly shapes I made myself, spanning a time period of seven years. I always aim for max. variety and it's a little bit annoying to see how they all tend to gravitate towards a certain style. I suppose that's inevitable.)

Thanks for the numbers and example images. Those figures do look pretty good for a body and yes, with proper LOD handling that would give decent performance while still giving a good amount of detail.

Those system heads look pretty good. They could use some more geometry around the cheek bones and the chin, if you ask me, but other than that they're cool.

  • Like 1
Link to comment
Share on other sites

4 hours ago, FinnfinnLost said:

They could use some more geometry around the cheek bones and the chin, if you ask me

Yes, those are the main areas that could obviously do with a few more tris.

The head and neck of the system body Aquila showed us account for 1,844 of those 7,186 tris which would explain why it's significantly higher visual quality than the rest of the body. Even if we double that number and add a few tris for the eyes we're still not at 4,000.

Add this to Ruth 2's 24,214 and let's round it up to an even 30,000. I think that's my final answer to the question in the thread's title.

Edited by ChinRey
  • Like 1
Link to comment
Share on other sites

I know people don't like me to bring this as an example, but look at Utilizator's Avatar 2.0.

It clocks at 34K head included, and has what i'd consider an excellent topology.

Edited by Kyrah Abattoir
  • Like 1
Link to comment
Share on other sites

20 minutes ago, Penny Patton said:

On an alt I've been using the Kemono body+Rei's Chest2 mod+Glutz Leg mod+Utilizator's Normie head+finger and toenail mods+prim eyeballs.

All that is a grand total of 39k triangles. 

 

kemono mod.jpg

Looks pretty good and has lots of potential for customization if you ask me. Although I don't particularly care about torpedo breasts.

There should really be some generous hard limit on tris for an avatar.

Link to comment
Share on other sites

11 minutes ago, FinnfinnLost said:

There should really be some generous hard limit on tris for an avatar.

And a limit on textures. I'm seeing more and more avatars pushing 500mb-1GB of VRAM. Some of us were begging LL for such limits back before mesh was released to the main grid. The Lindens at the time were confident content creators would be sane and responsible.

Our mutual acquaintance in the other thread illustrates how wrong they were.

ETA:

And the breasts are less torpedo-like out of the shirt, but had to keep it   forum friendly.

Edited by Penny Patton
Link to comment
Share on other sites

On 4/4/2021 at 4:51 AM, ChinRey said:

But that includes five separte meshes for different feet angles and only one is render at any given time. Ideally you'd have separate pairs to wear for each angle of course but it shouldn't matter much for the actual render cost - although it does increase the streaming cost quite a lot. And just like the hands, it's the LoD bugs that ar the problem, not the tri count in itself.

I'm not sure what you mean there, we don't really have a "null" material in SL so i'm pretty sure all the hidden parts are, (even if alpha masked which will drop the rendering load) at least undergoing skeletal transform, which is quite costly.

Link to comment
Share on other sites

On 4/11/2021 at 10:05 AM, Kyrah Abattoir said:

I'm not sure what you mean there, we don't really have a "null" material in SL so i'm pretty sure all the hidden parts are, (even if alpha masked which will drop the rendering load) at least undergoing skeletal transform, which is quite costly.

I don't know how much load those superfluous alpha'ed out extra body parts (and hair styles) cause so I decided to gove them the benefit of doubt. Does anybody have any reliable data about it?

Btw, I brought the question about tri count for mesh bodies up with a good friend of mine who is both a very seasoned professional game asset modeller and familiar with SL. He insisted that 10,000 is enough, arguing (quite correctly) that the main issue with the old system body isn't the tri/vertice count but the rigging. He did show me a couple of body parts he had made and they were quite convincing but it didn't include any of the parts where I thought there would be the most need for extra tris so I'm not quite sure.

Even so, my Ruth 2 based estimate of 30,000 and Penny's Kemono based 39,000 are way too high. Those two bodies are of course far more performant than any of the major brands but they're still not anywhere near being optimized.

How about 15,000?

  

On 4/7/2021 at 12:23 PM, Penny Patton said:

And a limit on textures.

Yes but when it comes to mesh bodies, texturing is out of the mesh modeller's hand. You can't blame them for 1024x1024 nail textures and such.

Edited by ChinRey
  • Haha 1
Link to comment
Share on other sites

2 hours ago, ChinRey said:

Btw, I brought the question about tri count for mesh bodies up with a good friend of mine who is both a very seasoned professional game asset modeller and familiar with SL. He insisted that 10,000 is enough, arguing (quite correctly) that the main issue with the old system body isn't the tri/vertice count but the rigging. He did show me a couple of body parts he had made and they were quite convincing but it didn't include any of the parts where I thought there would be the most need for extra tris so I'm not quite sure.

The system body/head has very noticeable problems with areas that should be a smooth curve and yet you can clearly see that the curve is made up of flat edges.  There are areas on the face that are very blocky too.  So to say the main problem is not with those but the rigging ignores the fact these blocky areas are the main reasons people went to mesh head/bodies in the first place.

So I am pretty sure that analysis is wrong.

  • Like 1
Link to comment
Share on other sites

38 minutes ago, Gabriele Graves said:

The system body/head has very noticeable problems with areas that should be a smooth curve and yet you can clearly see that the curve is made up of flat edges.  There are areas on the face that are very blocky too.  So to say the main problem is not with those but the rigging ignores the fact these blocky areas are the main reasons people went to mesh head/bodies in the first place.

So I am pretty sure that analysis is wrong.

The simplistic rigging is certainly a major issue and there's a lot of room for improvement there.

Apart from that I suppose you understood from my post that although I wouldn't say he was wrong, I do believe he exaggerated a bit. Even so, going from 7,000 to 10,000 means 3,000 more tris. You can smooth out a lot of curves with that. Also, I've only had a quick look at the system avatar mesh but there were a few normals that looked a little bit suspicious at first glance. Maybe some tweaks there would help smooth things out? Then there are the shape adjusters and the skin texture of course, both crucial factors for achieving a smooth looking face.

Link to comment
Share on other sites

3 minutes ago, ChinRey said:

The simplistic rigging is certainly a major issue and there's a lot of room for improvement there.

Apart from that I suppose you understood from my post that although I wouldn't say he was wrong, I do believe he exaggerated a bit. Even so, going from 7,000 to 10,000 means 3,000 more tris. You can smooth out a lot of curves with that. Also, I've only had a quick look at the system avatar mesh but there were a few normals that looked a little bit suspicious at first glance. Maybe some tweaks there would help smooth things out? Then there are the shape adjusters and the skin texture of course, both crucial factors for achieving a smooth looking face.

I do understand what you are saying.  I suppose that on the back of that exaggeration, it doesn't convince me that 10K is an less of an exaggeration that's all.

I think someone eventually is going to have to come up with a mesh avatar that either proves or disproves the theory that 10K or maybe 10-15K would be enough to satisfy mainstream mesh avatar users.  Did someone say that Ruth 2.0 is 30K?  Maybe someone could take that, I believe it is opensource, and optimize it to 10K so that there is a point of comparison that proves at least the reduction from 30K to 10K produces indistinguishable results.  That seems to be a worthwhile exercise and getting a point of comparison with a mainstream body seems like it wouldn't be feasible.

My feel though is still that wouldn't be the case and that's because "good enough" is very subjective when it comes to how things look.  I would be glad to be wrong though.
The idea of having even 10-15K for an avatar that is indistinguishable in visual quality from Maitreya Lara,  is certainly an enticing idea but ultimately just talk until proven and realized.
 

Link to comment
Share on other sites

This is a bit off topic but not completely. I was looking through my SoaS inventory and stumbled across these two handsome NPCs guys I picked up from some OAR years ago.

bilde.thumb.png.0b6b70c3ac16cdd9c7e95cf2fd9333e4.png

Here are the stats. Can you tell which is which?

bilde.png.139693dbf51a6ba42f0ee2392af013e4.png

Edited by ChinRey
Link to comment
Share on other sites

2 minutes ago, Gabriele Graves said:

The idea of having even 10-15K for an avatar that is indistinguishable in visual quality from Maitreya Lara,  is certainly an enticing idea but ultimately just talk until proven and realized.

I'd love to do that but I would have to learn how to rig and I don't really have time.

30,000 is definitely possible. I think we've proven that beyond any reasonable doubt now. The more I think of it and look at it, the more convinced I am it's possible to go even lower and still get a nice and smooth lookign avatar. How much lower, I dare not say for sure.

Link to comment
Share on other sites

I could try and guess but no, it isn't obvious.  However we only have just one static image to go on, cannot see everything and both are pretty rough and blocky looking.  I would think the real test is when you want to make an avatar that doesn't look blocky at all.  It is interesting though how bad one of them is compared to the other.  I guess that right hand avatar matches the left hand stats.

Link to comment
Share on other sites

3 minutes ago, ChinRey said:

I'd love to do that but I would have to learn how to rig and I don't really have time.

I'm not suggesting that you should but like I said, someone could certainly create a basis of comparison if they could and they did.

3 minutes ago, ChinRey said:

30,000 is definitely possible. I think we've proven that beyond any reasonable doubt now. The more I think of it and look at it, the more convinced I am it's possible to go even lower and still get a nice and smooth lookign avatar. How much lower, I dare not say for sure.

Sorry but if you think that the Kemono body proves that 30K looks as good as a M. Lara, you just don't see what myself and the rest of the market does.  This is what I mean about the subjective difference between what one group of people would be happy with versus another.  I remain to be convinced it would be enough to recreate M. Lara indistinguishably from the 176K it is today.

Link to comment
Share on other sites

2 minutes ago, Gabriele Graves said:

Sorry but if you think that the Kemono body proves that 30K looks as good as a M. Lara, you just don't see what myself and the rest of the market does. 

The thing that's not obvious about the kemono avatar I shared is that there's a whole bunch of superfluous triangles hidden from view. The upper torso and nearly the entirety of the legs are are duplicated by the mods I'm using. If I were to take the time to rip apart the base kemono and remove the redundant pieces then the tri count would be even lower. I also included the eyes, head and both finger and toenails in the total triangle count which probably aren't included when we talk the total triangles in the Maitreya body. Maybe later when I can log in I'll tear down a copy of the kemono to see what the triangle count is with just the visible body parts.

 Speaking of, when we're talking the Lara body being 176k tri, what exactly are we talking about? Are hands and feet included in that?

Link to comment
Share on other sites

Yes they are now.  The body is all-in-one and fully intact except head as of v5.0.  We are on v5.3 currently but in v4.x the hands and feet were separate.  The 176K is for v5.3

Forgot to add, that's without wearing any of the extra applier compatibility layers too.

Further info:

Nails: 5 sizes/shapes of nails on hands
Neck: two different neck sizes
Feet: 5 different foot positions
HD Nipples: Done as a extra tiny mesh layer over each nipple (TBH I could care less about this).
Alpha cuts: 208 different cuts - not sure if this affects number of triangles at all though.

These extra things mostly add extra mesh to achieve with the not-selected options being made invisible.

Edited by Gabriele Graves
Link to comment
Share on other sites

You are about to reply to a thread that has been inactive for 110 days.

Please take a moment to consider if this thread is worth bumping.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...