LL: Mesh is useless?

[Pamela Galli]

I have a 608 face bedpost -- a pretty plain bedpost.  It uploads with a LI of 8 on the most recently updated beta viewer. That's 32 for just four bedposts, not counting any other parts of the bed. I will not bother making the rest of the bed. There's just not much market for a 200 prim bed, no matter how pretty it is.

[Braydon Randt]

I was all set up to say something positive on this ..... and to be honest ,  im finding it hard to,  as I work in quads like normal folk ...  yet SL doubles face values as it gets converted to triangles because yanno ...  thats SL avatar animates so well with triangles dont it.  plus ...  I modelled something  and uploaded it  less than a weekago and it came to L11 and i took it and removed vertices ...  and today its asking for L$18  so ..  I will put this nicely..  LL sort this out please,  you want SL to attract people,  yet you bring in meshes that cant (or wont allow it to) use normals  so a 4k texture that you have compressed to crap for normals cant be used so reasonble attractive meshes are hammered.  I have become over obcessed with making every model i make look like a low poly WoW character.

So Pamela,  I can only offer that ....  errrr  I cant offer anything as at 608 Faces even as triangles is 1016 not even a 10L$ sculpty that has 1024.

 

[Pamela Galli]

Thanks for your reply Braydon.  I don't know if the algorithm has changed or what -- I have been getting a lot of wildly varying PEs on upload, but I don't think I can get this 8 PE bedpost down to the 1 prim that a sculpt takes.  The mesh may look better, but not 7 prims better.

 

Mesh has the potential to transform SL but if there is little market for it, it will have very limited use.

[Aquila Kytori]

hi, u need to perhaps make ur own Physics , for example a simple box shape surounding ur bedpost will, i guess  give it a PE / PI of 2. 

[Chosen Few]

I'm curious about two things.  First, why address your post to LL, when Lindens don't read these forums?  Second, instead of crying doom and gloom, why not just ask for help? Help is what we're all here for, after all.  Simply complaining accomplishes nothing.

There's no reason a bed should cost 200 prims.  And the good news is there are tons of things you can do to lower the land impact.  Don't blame the system, just because you don't yet fully understand how to use it.  Instead, learn, learn, learn, just like many of us already have.

First, how did you go about creating the model?  Did you UV it as efficiently as possible?  Did you create well-made LOD versions?  What did you use for the physics mesh?

Here's a bedpost I made just now:

I'm pretty sure you'd agree the high LOD version is about as detailed as any bedpost really needs to be.  It's 656 tris, a hair more than yours.  The medium LOD is 464 tris, and the low is 140.  I didn't bother making a lowest LOD, since no one's ever going to see a bed from far enough away that that will ever matter.  I'll just let SL reduce it to whatever it wants, at the lowest possible setting.

I used cylindrical mapping for the UV layput, so it's all a single shell.  In SL, that translates to zero duplicate vertices, which keeps the download cost low.

Now here it is in-world:

As you can see, it's got a land impact of 2.  That's due to the physics weight of 1.9.  I used the low LOD for the physics mesh, and let SL simplify it.  Had I made a simpler one myself, I could well have dropped the physics weight to under one prim.  I'm not sure that would have lowered the download weight, though, so the LI probably still would have remained at 2.  But 2 is a hell of a lot better than 32, obviously.

With four of these, plus a headboard, footboard, mattress, and frame, the entire bed would likely come out to somewhere around 15-20 prims, tops.

 

As I said, there's no need for a bed to be 200 prims, and there's no need for a bedpost, even an ornate one, to be 32 prims.  Even one ten times as complex as the one I made would only be 20 prims.  If a 600-polygon one is coming out to 32, it's a clear cut case of user error.

[Chosen Few]

---

Braydon Randt wrote:

I work in quads like normal folk ...

---

There's no need to work entirely in quads, unless you plan on stepping up the poly count in a program like Z-brush or Mudbox, for high-detail texture baking purposes, or normal mapping (which SL can't even utilize).   And even if you are doing that, it's always a good idea afterwards to go back to your original model, and eliminate triangles wherever you can.

Modeling for any realtime environment is about utilizing the least amount of polygons possible, always.  If you've got extra tris or extra quads in there that are unnecessary for describing the shape of the model, get rid of them.

 

---

Braydon Randt wrote:

yet SL doubles face values as it gets converted to triangles because yanno

---

Every realtime engine on the planet deals with triangles, rather than quads.  This is not an SL thing.  It's a basic computing thing.

 

---

Braydon Randt wrote:

thats SL avatar animates so well with triangles dont it.

---

The SL avatar is made of triangles, just like every other polygonal model, in every other realtime simulation on Earth.  A good many of the triangles happen to be organized to fit together into quads, which is a good thing, of course.  But not all of them are, nor do they need to be.

 

---

Braydon Randt wrote:

I modelled something  and uploaded it  less than a weekago and it came to L11 and i took it and removed vertices ...  and today its asking for L$18  so ..  I will put this nicely..  LL sort this out please,

---

It's not LL that needs to sort it out.  It's YOU that needs to learn how to use the system.  When you deleted vertices, you probably triggered a change somewhere in the mesh structure or the UV layout that ended up increasing certain costs, while lowering certain others.  You have to take all factors into account.  If you don't know what the factors are, either ask here, or go to the wiki, and read, read, read.

One should never model for any platform without knowing all the technicalities involved.  Try modeling for a PS3 game some time.  Sony sends you documentation a mile thick, and you have to know it all before you can even begin.  SL is quite tame by comparison.

 

---

Braydon Randt wrote:

meshes that cant (or wont allow it to) use normals

---

Careful with your terminology.  SL meshes absolutely do use normals, just like all meshes in all simulations.  If they didn't, you wouldn't be able to see them, as light would not reflect off of them.

SL suports the use of hard and soft edge normals, and mono-directional face normals.  I would have been unable to create the combination of smooth surfaces and hard edges without that support.

What it doesn't yet support is normal mapping.  It will eventually, but not yet.  Don't hold your breath waiting, though, as the feature could be tomorrow or next year or five years from now.  We have no way of knowing when it's coming.

[Aveline Stein]

---

Chosen Few wrote:

One should never model for any platform without knowing all the technicalities involved.  Try modeling for a PS3 game some time.  Sony sends you documentation a mile thick, and you have to know it all before you can even begin.  SL is quite tame by comparison.

---

 

Wow i would very much love to have better documentation on SL Mesh as oposed to the 'there you go, and let the community figure it out' approach.

I am far from being a professional myself, but have doodled with other engines and every single one of them has better documentation than SL has when it comes to mesh and importing things.

Considering that SL is aimed mainly at hobbyists, one would think that they would at least put out some guidelines and video tutorials for good practices. Posting code snipples and formulas on the wiki may be great to some people, but i doubt it's helpful to the majority.

 

Looking at the forums there seem but a few people who really fully understand what is going on, and the amount of bugs and oddities are not helping to understand it better in the least.

Alot of the whole mesh process is outright confusing, uploading the exactly same file yields different upload costs at a different hour.

PE of an object can change from 200 to 0.5 just by using different LOD settings (and more).

No documentation on errors the viewer gives (MAV_BLOCK_MISSING), greyed out upload button with no meaningful error message to work with, and in some cases red colored vertices or grey edges in the uploader preview i could find no info on.

All this happening on models that pass through Mayas mesh 'cleanup' function with zero errors.

For me personally, documentation is just plain lacking, when in fact it should be *better* than what is available for indie and professional engines due to the less expierenced target audience.

 

If LL are truely hoping to get mesh fully accepted on a wider scale and to replace the awefully inefficient sculpties they should work on this.

And before someone jumps me, i do not think people want to be told how to use their tools, but would rather know about SLs unique requirements and kinks.

The problem isnt unique to mesh however. I wonder how many prim and sculptie builders are aware that they are now able to lower the landimpact (and thus decrease the load on sims) on their builds by utilizing the mesh impact calculations for their prims and sculpts. My guess is 'not many' because unless they read the wiki article very carefully and experiment, or run into oddities when building they will never know.

The idea of a world where everyone can create is a great one, but without guidance and best practices that are easy to find and understandable, it will take a long time for SL users as a collective to produce non-laggy, conforming items.

[Pamela Galli]

In answer to your two questions: 1) Lindens do read the forums, and even post in them, including this one. 2) I did not ask for help because this is not a simple matter of me needing advice or help -- it has to do with the wildly fluctuating upload process which produces a different LI for the same mesh every time it is uploaded. As far as complaining, I have a very long history of complaining loudly (in the forums and in email) and starting JIRAS, about anything and everything that affects my business (which happens to be my full time job). I don't plan to stop because you don't approve.

 

I don't know why you assume I have not "learned, learned, learned" anything. Here is a chair I made (with help from Nacy) that is 5 PE/LI with two scripts in it:

 

 

Everything was uploaded with the exact same parameters as the bedpost. Each arm is 746 faces, so comparable to the bedpost. And yes, getting a LI that low was a matter of rolling dice, as the LI was all over the place depending on the configuration of the stars at the time, I am guessing.

 

My bedpost is 8, not 32 prims, btw. I was planning on using 4 of them.

 

 

 

[Pamela Galli]

Aveline, I could not agree more. Much of the official (or unofficial) information I have been able to find is out of date. If there are any plans to write a comprehensive guide to "Mesh in Second Life" I have not heard of them.  I would like to know why my chair with a script in it is 5 LI and my one bedpost (unscripted!) using the same upload parameters is 8.

[arton Rotaru]

If you would provide any information which weight is 8 (download, physics), we probably could tell you what's wrong with it.

I suspect you haven't specified a physics shape during upload. In this case the system creates a physics shape around your mesh on it's own. If the mesh is round, such a physics shape can be rather expensiv. Always upload your own physics shapes with your mesh. Never let the system create one for you. A simple box shape as physics results in a physics weight of 0.36.

If it is the download weight which is 8, well Chosen showed how it can be done cheaper..

[Pamela Galli]

---

arton Rotaru wrote:

If you would provide any information which weight is 8 (download, physics), we probably could tell you what's wrong with it.

I suspect you haven't specified a physics shape during upload. In this case the system creates a physics shape around your mesh on it's own. If the mesh is round, such a physics shape can be rather expensiv. Always upload your own physics shapes with your mesh. Never let the system create one for you. A simple box shape as physics results in a physics weight of 0.36.

If it is the download weight which is 8, well Chosen showed how it can be done cheaper..

---

Last time I specified a physics shape during upload, the upload button was grayed out.

However that is beside the point, which is, as I have stated several times: Why, using the same upload parameters, can I have a chair with a PE of 5, but the bedpost is 8?

I just uploaded it again. This time the LI is 6.8. Same mesh, mind you. I knew it would be lower because all the numbers on the first page were lower -- I don't remember them all, but the lowest one was 44 before -- now it is 18.  I did not touch these numbers, and here is how they looked:

 

Here are the caluclations:

[arton Rotaru]

Yeah, you use the Hi LOD in the first 3 slots. At least you should lower the Low LoD to some degree. I think Drongle has stated it more than once, if you use uploaders LODs, those can vary everytime you upload a mesh. The LOD creation is non deterministic. It will create the Lods on a more random basis. Which, of course results in different PE results.

I for one, I never use any of the importers LODs. I always make all 4 LoD myself and upload those, with a selfmade physics mesh as well. I never see such big differences between uploads.

 

About the greyed out button, see http://community.secondlife.com/t5/Mesh/Can-t-upload-when-using-physics-from-file/td-p/1265005

[Pamela Galli]

Those are the numbers that were in the uploader; I didn't enter them.  For my 5 prim chair (went to 9 with a script but I brought it back down to 5 with advice from the kind ppl in this forum), I used the model for the first 3 LODs and 25 for the last one. Did that with every mesh, just as I did with the post.

As I mentioned, when I used a physics shape I made -- the upload options were grayed out. That may have to do with me having a Mac, I don't know. I am using the development viewer, and have never used anything but that.

[arton Rotaru]

Oh Mac. k

Well, as I mentioned. I would highly recommend to create all your LODs by yourself. Think about, to show a bed post in lowest LOD all you need is a cubic shape with 8 triangles. Top and bottom open. This way you have still a "post" and way less triangles in lowest LOD than 18, which gave you the importer the last time for your lowest LOD.

Apply the same method for the low LOD, with a few more triangles, and your bed post will be 1.x PE.

[arton Rotaru]

Actually, a lowest LOD out of 6 triangles would be good enough.

For every single tirangle you remove from the lowest LOD, you can add 4 tris to the highest LOD

[arton Rotaru]

---

Pamela Galli wrote:

Those are the numbers that were in the uploader; I didn't enter them.  For my 5 prim chair (went to 9 with a script but I brought it back down to 5 with advice from the kind ppl in this forum), I used the model for the first 3 LODs and 25 for the last one. Did that with every mesh, just as I did with the post.

As I mentioned, when I used a physics shape I made -- the upload options were grayed out. That may have to do with me having a Mac, I don't know. I am using the development viewer, and have never used anything but that.

---

Wait, just to get this right. You say when you select your mesh, the importer shows you the same numbers in Medium and Low LOD as in Hi LOD by default?

[Pamela Galli]

---

arton Rotaru wrote:

---

Pamela Galli wrote:

Those are the numbers that were in the uploader; I didn't enter them.  For my 5 prim chair (went to 9 with a script but I brought it back down to 5 with advice from the kind ppl in this forum), I used the model for the first 3 LODs and 25 for the last one. Did that with every mesh, just as I did with the post.

As I mentioned, when I used a physics shape I made -- the upload options were grayed out. That may have to do with me having a Mac, I don't know. I am using the development viewer, and have never used anything but that.

---

Wait, just to get this right. You say when you select your mesh, the importer shows you the same numbers in Medium and Low LOD as in Hi LOD by default?

---

Exactly. That's why I took the pic.

[Pamela Galli]

---

arton Rotaru wrote:

Oh Mac. k

Well, as I mentioned. I would highly recommend to create all your LODs by yourself. Think about, to show a bed post in lowest LOD all you need is a cubic shape with 8 triangles. Top and bottom open. This way you have still a "post" and way less triangles in lowest LOD than 18, which gave you the importer the last time for your lowest LOD.

Apply the same method for the low LOD, with a few more triangles, and your bed post will be 1.x PE.

---

Let me correct myself -- the bedpost was originally 11 PE.  I lowered the triangles to something like 4, and brought it in a 8.

[arton Rotaru]

Well, that should only happen the second time you upload the same mesh. It uses the yourmeshname.slm file than. If this happens with a mesh you never uploaded before, this must be some kind of Mac viewer bug. Search your folder where you have your meshes, and delete the corresponding .slm file, and see if the importer shows still the same numbers for the first 3 LODs.

Though, I can't say it often enough, make all your LODs by yourself. It's more work, yes, but it's worth it.

[Pamela Galli]

I will do what I have to do, of course, even if it means making 10 LOD models - - but until now, except for the variability, I was not having a problem winding up with low PEs with the LOD models I have been using.

[Gaia Clary]

---

Pamela Galli wrote:

Those are the numbers that were in the uploader; I didn't enter them.  For my 5 prim chair (went to 9 with a script but I brought it back down to 5 with advice from the kind ppl in this forum), I used the model for the first 3 LODs and 25 for the last one. Did that with every mesh, just as I did with the post.

---

The Mesh importer will by default reduce the automatic calculated LOD's for each level. If it behaves like you describe (reuse LOD3 for LOD2 and LOD1) then you might look for a file with name "{YourUploadFileName}.slm" besides the "{YourUploadFileName}.dae" that you import. If such a file exists, this contains the "last used import setup" for the associated object and that presets your importer when you upload the next time.

You can reset the Importer to get rid of this preset.

Some more remarks on other topics raised in this thread

Regarding usage of LOD

Are you aware that the most important LOD is LOD2 and not LOD3? This is because of the LOD transition points. LOD2 covers the mid range of distance where the object can still be seen with lots of detail. The transition point to LOD1 is typically at a distance where your object no longer is the most important element of the scene.

If you make your LODS carefully, then you won't even notice the transition points. I have found the following rule of thumb usefull for my stuff:

 

1. Reduce LOD3 as much as possible such that you are at the point where a closeup is still looking good, but not necessarily brilliant. Use Tris if they help to accomplish this task. (You can see in my shoe quest for example how Tris can help to maintain the shape while reducing polycount dramatically)
2. Try to bake details into textures even for LOD3.
3. Reuse LOD3 on LOD2. I can do this ONLY(!) when i have already optimized LOD3. Otherwise this makes not much sense.
4. Get LOD1 realy down. Remind that the object is normally far away at the LOD1 transition point.
5. Get LOD0 radically down. Often a very simple shape is sufficient. Sometimes a billboard will do it. Note that LOD0 is the most important LOD regarding streaming costs for small objects like bed posts.
6. If your object does not need physics, then create a one triangle physics mesh. This dramatically reduces physics costs. I believe you also can set physics to none for child objects in link sets. That should also do the trick.

Here is a shoe which needs 978 tris on LOD3/LOD2 and 526 Tris on LOD1 (using Tris where reasonable):

Regarding non deterministic behaviour of LI and PE:

When i keep the LOD Generator completely inactive (i create all LOD's by myself, including the physics shape) then i always get the same LI/PE. I have never seen any random change in the face count.

Regarding increase of PE when vertices are removed:

When you remove vertices, then maybe you also have changed your seam definitions? This definitely has an influence on the PE because the number of vertices in the UV map counts too.

[Chosen Few]

---

Pamela Galli wrote:

In answer to your two questions: 1) Lindens do read the forums, and even post in them, including this one.

---

Lindens are typically too busy doing their jobs to read forums that are intended for users helping users.  From time to time, certain individual Lindens happen to be kind enough to volunteer their time to post, upon having been directly asked to do so by an interested resident.  Beyond that Charlar regularly posts in the usergroup thread, with meeting notes, announcements, and such .  But I haven't known any Linden to go reading through the forums as a matter of course, at least not since circa 2005, when the community was small enough that they could employ a dedicated staff to do that.

The point was if you want to reach them directly, the forums are not the way to do it.  On this particular subject, even if you did go through proper channels, it's unlikely you'd get much of value as a response, anyway.  Just because the Lindens built the system in no way means they're the best people from whom to get advice on how to use it. 

Think about it.  If your golf swing were unpredeictable, exactly how much good do you think it would do you to post a sign up at the local course, stating "Wilson:  Nine iron useless?"   Wouldn't it be far more sensible just to ask fellow golfers for advice on what YOU could be doing better, rather than blame your woes on the makers of your tools?

The information you needed to solve your particular problem in this case was advice on how improve your own modeling technique, your own approach to modeling for SL specifically.  It is we, your fellow users, who are are best equipped give you that.  Leave the Lindens to do their jobs, and ask the rest of us for help with usage.  That's what we're all here for.

Had your problem actually been an issue with the system, rather than a symptom of your own lack of understanding of how it works, then by all means, the right thing to do would have been to alert the Lindens.  But as I'm sure you know, the forums aren't the appropriate venue to do that either.

 

---

Pamela Galli wrote:

I did not ask for help because this is not a simple matter of me needing advice or help -- it has to do with the wildly fluctuating upload process which produces a different LI for the same mesh every time it is uploaded.

---

Yes, it is absolutely a simple matter of needing advice and help.  Clearly other people besides you are able to upload without such wild fluctuations, right?  So doesn't it stand to reason that perhaps some of us know something you don't?  I assure you, we're not using magic, and there's no evil overlord out there who thought it would be really funny for you to receive the crappy uploader while we got the super secret good one.  We've just spent a little more time discovering what works than you seem to have at this point.

Here's the crux of it.  If you're seeing the numbers fluctuating in the manner you described, the ONLY possible reason for that is that you're too reliant on the automation to do the work for you.  You should be controlling the process yourself, rather than pushing a button and hoping for the best.  That means making your own LOD versions, along with your own physics mesh, and intelligently considering the best way to structure each of them, every single time you make a mesh model.  Those of us who actually do that never experience the apparent randomness that you've thus far implied has been unavoidable for you.

In other words, the problem here is user error, not system error.  And that's terrific, because it means YOU can fix the problem, quite easily, and not have to suffer any further from it.  All you need to do is follow the advice that those of us who know how to avoid the problem are offering.

 

---

Pamela Galli wrote:

As far as complaining, I have a very long history of complaining loudly (in the forums and in email) and starting JIRAS, about anything and everything that affects my business (which happens to be my full time job). I don't plan to stop because you don't approve.

---

It's not about whether or not I approve.  It's about what's the wisest strategy to employ, if your goal is actually to improve your situation.  If all you want to do is hear your own voice every time you're upset about something, that's your prerogative, I suppose.  Me, I'd rather focus simply on results and how-to's. 

When you run into trouble with anything, you can either shout, "This doesn't work!", or you can ask, "What can I do to make it work?"  The former is never productive.  The latter almost always is.

 

---

Pamela Galli wrote:

I don't know why you assume I have not "learned, learned, learned" anything.

---

I never said you haven't learned anything at all.  If you plucked that notion from anywhere, I might suggest you look toward your own thoughts and feelings as the source, rather than into anything I said.

My point was simply that you haven't yet learned enough to be able to control the outcome of your uploads.  That's evidenced in spades by your own description of the obstacles you've encountered. 

You've got two choices at this point.  You can acknowledge that fact, and seek the requisite knowledge to solve the problem, or you can just continue bitching about it.  I will continue to present the information you need to know (and so will other helpful souls, many of whom have already begun doing so in this thread).  You can choose to accept it or not.

 

---

Pamela Galli wrote:

Here is a chair I made (with help from Nacy) that is 5 PE/LI with two scripts in it:

Everything was uploaded with the exact same parameters as the bedpost. Each arm is 746 faces, so comparable to the bedpost. And yes, getting a LI that low was a matter of rolling dice, as the LI was all over the place depending on the configuration of the stars at the time, I am guessing.

---

From reading this, I can tell right away what the problem is.  You're only uploading one LOD, right?  And you're letting the system create everything else?  That's mistake number one.  If you want the outcome to be predictable, you must control it yourself.  Don't surrender control to the system.

I see your later posts go on to confirm my hypothesis, and I see that others have stepped up to offer the very same advice in response.  Make your own LOD's and your own physics mesh, and the problem will go away.

 

---

Pamela Galli wrote:

My bedpost is 8, not 32 prims, btw. I was planning on using 4 of them.

---

Ah, sorry about that.  I was late to meet some people when I wrote that, and was in a hurry to get out the door.  Apparently, I fumbled the two numbers in my haste.

What threw me was your line about the final bed being 200 prims.  I figured you'd arrived at that estimate because the four posts would equal 128, at 32 each, and that the rest of the bed would constitute the other 72.  That's about the right ratio, even with the numbers themselves being so absurdly high.

With 8 prims per post, I'd expect the whole bed to come out at somewhere around 40-50.  I'm not sure where you got 200.

At my example of 2 prims per post, I'd estimate the finished bed at a maximum of 15-20, as I said.  Realistically, it would likely be well under that, since the bed as a whole would only need a single physics mesh.  This encompassing singular mesh would require far less triangles than would the sum of individual physics meshes for each part.

 

---

Pamela Galli wrote:

I would like to know why my chair with a script in it is 5 LI and my one bedpost (unscripted!) using the same upload parameters is 8.

---

The answer is simple.  You allowed the system to create the LOD's and physics mesh for both objects.  All other things being equal, the chair, which is "square-ish" can generate a simpler physics mesh than the post, which is "roundish", with any given setting.

 

---

Pamela Galli wrote:

Last time I specified a physics shape during upload, the upload button was grayed out.

---

Did you apply the same material(s) to physics mesh that are on the visual mesh?  There's a known bug right now which imposes that requirement.  All meshes in a model must have the same material(s) on them.  This was meant to apply to the LOD's only, but the physics mesh got lumped into the mix accidentally.  The Lindens have stated intent to fix that, but in the mean time, it is what it is.

 

---

Pamela Galli wrote:

I just uploaded it again. This time the LI is 6.8. Same mesh, mind you. I knew it would be lower because all the numbers on the first page were lower -- I don't remember them all, but the lowest one was 44 before -- now it is 18.  I did not touch these numbers, and here is how they looked:

---

Again, the more you rely on the automation, the more unpredictable the results will be.  It's not designed to produce target numbers repetitively.  It interprets the mesh from scratch each time, and makes fresh judgments about how best to handle it each time.

If you want full control over the outcome, make your own LOD's and your own physics mesh, and leave the automation out of the equation.

 

---

Pamela Galli wrote:

I will do what I have to do, of course, even if it means making 10 LOD models - - but until now, except for the variability, I was not having a problem winding up with low PEs with the LOD models I have been using.

---

Fantastic.  I'm thrilled to know you're open to doing what you need to do to make this work.

I do have to say I wonder why, if you really weren't having a problem arriving at low numbers previously, did there appear to be so much venom aimed at that particular subject in your original post.  The message you seemed to be trying to send was you felt mesh as a whole was useless because it carries higher land impact than you felt was practical.  If the high land impact problem was limited to just this one model, why not simply ask for help with the model itself, instead of throwing all previous experience out the window?

You can feel free to interpret that as a rhetorical question.  The answer doesn't really matter to me, as long as you now understand that the problem is fixable by you.

 

 

I had planned to do a nice little writeup of good LOD strategy here, but Gaia beat me to it, and her post is excellent (as always), so I won't bother restating the same information.  Hopefully, between her general points, and my earlier specific example, there's enough there for you to see how you can best control all this.  But if you need further help with it, by all means feel free to keep asking, and I'll be more than happy to add.

[Pamela Galli]

"From reading this, I can tell right away what the problem is.  You're only uploading one LOD, right?  And you're letting the system create everything else?  That's mistake number one.  [...]

I see your later posts go on to confirm my hypothesis, and I see that others have stepped up to offer the very same advice in response.

[...] Here's the crux of it.  If you're seeing the numbers fluctuating in the manner you described, the ONLY possible reason for that is that you're too reliant on the automation to do the work for you.

[...]

Again, the more you rely on the automation, the more unpredictable the results will be.  It's not designed to produce target numbers repetitively.  It interprets the mesh from scratch each time, and makes fresh judgments about how best to handle it each time.

If you want full control over the outcome, make your own LOD's and your own physics mesh, and leave the automation out of the equation."

As I said earlier:  "For my 5 prim chair (went to 9 with a script but I brought it back down to 5 with advice from the kind ppl in this forum), I used the model for the first 3 LODs and 25 for the last one. Did that with every mesh, just as I did with the post."

(Then I amended that to say that when I was uploading the bedpost I lowered the last number to get the post from 11 to 8.)

 

 

 

 

 

 

 

 

[Pamela Galli]

Hi Gaia -- thank you for your reply.  I have watched your LOD video, and many others, and thank you for those, as well as your inworld group help :-)

I have not been using lower LOD models, but reusing the high LOD model for the top three levels, on every mesh. Until this bedpost I was able to get a pretty low LI -- I saw no reason to not use the best quality mesh for all but the last level, when I could get low LI as well as good LOD (see my 5 prim scripted chair).

However, the uploader did fluctuate for me, almost without exception, even though I was entering the exact same parameters each time. 

Why, I don't know. Apparently when others use models rather than letting the uploder generate LODs they do not experience this fluctuation.

 

 

 

 

 

[WADE1 Jya]

Mesh isn't useless, far from it! Mesh is the format the whole wide world uses for 3D design outside of SL. I'm finding great applications for my meshwork, it's all going wonderfully :matte-motes-sunglasses-2:

What I don't like is Lindens still allowing incompatible viewers to access the grid. Prevalence of incompatible viewers makes SL a fractured market. Also by default settings, lots of people still can't see mesh very well. More market fracturing, which is bad bad bad. Mesh should be full on & ready to go for everyone by default so we can all share the same experiences.