Gaia Clary

Gearsawe Stonecutter wrote: I now don't care as much to optimize low poly stuff since it offers no benefit to cost and only causes visual problems in the end. Sure it is nice of me to do so and the only selling point is it will render fast on your machine! From what i found regarding costs: Reducing the lowest visible LOD has high impact on resource costs. In fact it is the only way to get streaming costs down to reasonable numbers. I found that using a billboard-like approach for the lowest visible LOD gives very acceptable look, and reduces the Streaming costs. Please note that the lowest visible LOD varies with object size! For small objects up to a Bounding Box radius of ~5.43 meters (the distance between BBOX center and the BBOX corners) you can optimize the lowest LOD. If your meshes get bigger, then the lowest LOD is no longer used and the next higher LOD rules the price, and so forth until your mesh BBOX radius exceeds the size of ~43.4 meter where only the highest LOD is taken into account. Here is a table summarizing the numbers: BBOX-Radius (smallest) BBOX-Radius (biggest) max cubic BBOX (example) Lowest visible LOD Comment > 0 5.43 <6,6,6> LOWEST Costs can be reduce best with meshes <= this size 5.43 10.86 <12.5,12.5,12.5> LOW 10.86 43.4 <50,50,50> MEDIUM 43.4 > 43.4 HIGH Objects above this size only use highest LOD from any distance. So slicing up a huge object into many smaller parts and import them as a link set would be an option for making buildings etc. with less PE (i refer to Drongle's experiments on that) Always add a physics mesh, because the default setting uses the medium mesh as physics shape and creates a fitting Convex hull, which will probably yield very high physics costs. Better use the lowest LOD for physics or generate your own physics mesh. Optimize your object so that it looks good with medium LOD. Keep HIghes LOD for nice to haves but not realy needed details. This seems to make the overall behaviour of the objects much smoother regarding look from a distance. Give up Quad-centric modelling. It appears to me that Triangle centric modelling gives better ways for optimizing at least the streaming costs. Not that i say "only use triangles". Quads are good for texturing, but triangles are good for reducing. So you will have to find out what goes best with your particular object. But do not avoid triangles !!! Learn how to make optimal UV-maps. You can get rid of MANY faces if you make a proper UV-mapping. And you can enhance the look of your mesh significantly if you know how to unwrap. Collecting many mesh objects into one linkset reduces the PE per mesh.  If you physics cost for a link set rules the PE, then set children in your linkset to physics type NONE wherever possible. That may reduce physics cost again. I am not sure but from my very first experiments with upload costs i guess you can get back to reasonable costs per mesh when you upload a scene with many objects. So you can distribute the base cost of 150 L$ over all objects in the import. You later can unlink them again and use them as you like (but take care of point 3.) above) I am not filled with ehthusiasm when i see what is happening with mesh. But i would rather first check out the features before going into criticism. And it would realy help us to get some constructive input from the LL about where they want us to use mesh and in which way they want us to use mesh. And why they want us to do so. So maybe the intention behind all this is to move from single object productions to scenic development ? And maybe we will see entire scenes to get into the market and not single objects... Or partial scenes plus a bunch of buyable props intended to be linked into a scene ? All that seems to be supported by the new accounting. And please forgive me if i am too optimistic. But i must keep myself in a good mood somehow ;-)

IMHO what we see in the "new UI" can not realy be counted as improvement and i must say that i am very disappointed. I only can hope this is something intermediate which will be beautified soon (and the newly introduced errors get fixed). It looks to me like an interface that might focus on (and might be dictated by) technical relevancy. But customer needs are not well supported with that one. All was better 6 months ago. So what happened ? There is still an open Jira ( CTS-664 ) about UI and this jira was even requested by LL themself after closing CTS-611. So let us hope that at least a bit of what we collected in CTS-664 will eventually be seen in the final importer interface ;-( At least that... And BTW since Upload costs get significant now, you realy should add a popup saying (for example) "This upload will cost you 1250 L$. Are you sure you want to upload this now ? [YES] [CANCEL]"

Fyi: I added another Jira CTS-683 Can it be so complicated to add the Server Weight to the Debug informations in the SL-editor ??? Then we can at least see the 3 weight values which are used to derive the PE ...

For me it now looks more predictable and slightly better. Have you downloaded the newest viewer also ? Although i still do not understand this: Object SC PW SW PE comment 1 Bowling Pin 0.8 0.6 2 2 as expected 4 separate Bowling pins 3.2 2.2 8 8 as expected Linkset containing 4 Bowling Pins 3.2 2.2 3 7 I expected 3 here (note: SW= 1 + 4*0.5) I use the following abbreviations: SC (Streaming cost) PW (Physics Weight) SW (Server Weight) PE (Prim Equivalent) So the situation is slightly better compared to yesterday (where i got PE=9 for the linkset)

The Whole Collada exporter and importer seems to be under heavy development and there is still no finished version available. Parts of the exporter work however, especially static meshes could be exported since a long time, but with rigged meshes there have always been problems... Regarding static meshes, as far as i have seen, Blender can only export the entire scene, so i am not sure how i should create the 4 levels of detail meshes. I can not upload them all in one file, so do i have to place my 4 LOD meshes in 4 different scenes and export them one by one ? So up till today I always created my meshes with Blender 2.5 but use 2.49b for export to Collada.

Here are the first 2 video parts of the Kettle Quest. I am sorry to tell that the realy exciting parts will be part III and IV. But nevertheless i hope you will enjoy the first two also (they need a final polishing when mesh is ready to go, because there are still a few changes in Linden Labs queue before they are going to release). The videos are based on my article "The Kettle Quest") on the Machinimatrix Blog. Constructive Feedback seriously requested: Part I (Departure) Part II (Mimic Sculpty LOD)

So that would mean, mesh fails due to pricing, not due to technology ? Can it be that this is a wonderfull example for missfit to market ? Or the true plans are somewhat different. I tried to play the "if i where a Linden..." game. And this helped me to understand what might go on here. Of course my thoughts are highly speculative (i have no numbers on which i can back my arguments), so be kind with me and don't beat me for having the wrong ideas in my mind ;-) ( I am no merchant, no Landlord, and no Linden) So... When i where a Linden with a technical focus, then my main concern would be to reduce data traffic. Hence i would focus on getting the amount of data to be sent to viewers as low as possible. Now as a matter of fact, Sculpties need less data to be transfered. They have only one texture, the UV-layout is fixed, the number of verts is 1089 at maximum, they have no vertex normals to be transfered... So when i where a LL with technical focus, then i would propagate Sculpted Prims. When i where a Linden with a sense for market, i would ask myself which sort of stuff is mostly wanted ? Buildings ? Environments ? 3D sculpturing ? Clothing ? Avatar attachments ? ... I realy have no idea about the numbers, but it appears to me that Avatar's most interest is to look good... realy i do not know if this is true, but i think that many many users of SL do not have land. So if they ever buy something, then it is Avatar-attachments. Hence if i where a LL i would look deeply into that area. And i would see that meshes could give me a lot of benefits here regarding fighting lag... Ok, lets expand the game : When i where a Region owner, then i want to get my Land looking better than the land of my neighbour ... I want my house to look good, i want to give a nice environment to my friends and visitors, the best of the best is just good enough for me. If i had a shop, i want my visitors to talk about my shop and i want them to get back not only because my stuff is the best of all, but also my shop must look accordingly. If i had a mall, i want my renters to be happy and paying. So i might (hypotheticaly)... not care about prims. I have enough prims and i just use half of them anyways, so if i can get something that looks better, i might not care about the PE... So i will buy this phantastic stunning ultra cool looking statue of this old roman emperor, well ok it costs 200 Prims... but why not ? it looks soooo coooool .... I WANT it. If i where a poor land owner with a bit of land and limitted prim count, then i would want to limit my costs and rezz as many stuff as i can for as low prims as possible. OK i will buy this 1 prim full interior Sculpty, and this cool 1 Prim Piano and i am happy. I do not mind that it is not realy the nicest of everything. Its mine and its cool. so i can live with Prims and Sculpties. Hey, the builders are GOOOD in Sculpted Prims. So why bother with high cost stuff ? If i where here to just chat and look around whats going on, then i would be mostly interested in getting nice clothes, nice attachments... This nice looking jewelery, i would not care about putting on shoes with 60000 Triangles for each of them (because i do not know about such technical issues) , but if that other shoe looks so much better, i would prefer that one. i do not care about the material it is made off... Oh, myabe i WOULD care and heck, mesh is cool, so my mesh shoe looks better than my sculpty shoe. anyways, 200 Prims both ... so pfff... hey, the new one makes me walk faster... (less lag ;-) ) So i am happy with mesh (maybe without knowing) ... enough ... So i could imagine that mesh (on avatar) replaces high lagging sculpted attachments by lower lagging rigged attachments. And mesh (on ground) opens up different markets (architectural-, highest precision, machinima environments, Artwork) , and does NOT replace existing markets. Does that make any sense ?

Marcthur Goosson wrote: (I also found a bug: texture faces switching when changing LOD) I am not sure about that, but i always had the impression that you have to take care about the order in which the vertex groups for the different texture faces are exported. So you might have mangled something in your model ? So for primcount the sculptie wins but in my opinion mesh has far more options in texturing and building accurate models. If that is a plus in a prim based economy ? Sculpties might get a PE value to in future, HELP !! If Sculpties would get a PE in the future, wouldn't that be a content breaker ? Also i am afraid that many people would not care about "is it mesh or sculpty", but they would care about how many prims they have to spend to rezz an object. And i think this behaviour is very very legitimate. However i still think, that if done correctly, then a multi part mesh should become better compared to an equivalent sculpties-only creation regarding vertex-count, look&feel AND primcount. And the final numbers should make it possible to do that. Anything else would not make too much sense to me. My experiments so far show that this can be done wherever you use many sculpties (like 40 prim shoes, 200 prim hair, etc..) In such cases a mesh could reduce the PE weight dramatically. But wherever we talk about only a few complex objects, mesh fails and Sculpty will win. Some (one prim) objects even seem to be much easier to build with Sculpties. And i am pretty sure that some objects can be made with Mesh, where you have a lot of trouble if you wanted to do the same obejct with Sculpties. So at the end we will have good examples for every method (regular prim,sculpty,mesh). And i do NOT at all believe that Sculpties are "out of the game". They HAVE some nice features, if you like it or not ;-)

I found another issue with Mesh LOD. While making my video i tried to compare a Sculpty and a mesh with exactly the same internal mesh (i exported a sculpty as mesh). I also made the LOD's of the mesh according to how it is calculated for Sculpted prims. My expectation was that now both objects behave exactly identical. But they don't do that, meshes go to the next lower LOD much quicker. I have created a Jira for this ( https://jira.secondlife.com/browse/CTS-673 ) , because it looks wrong to me. if on the other hand mesh is supposed to change its LOD shape earlier than Sculpted prims, then we now have got another parameter where mesh is worse than sculpties. Here is an image comparison: Sculpty on the left, mesh on the right (For the handle i made a mistake on the LOD, hence the upper right and the lower left handle look a bit different. If i had done it correct, then they would look exactly the same). Nevertheless you can see on both images, that the Mesh objects turn to lower LOD much earlier than their Sculpty equivalents The upper image was made from about 40 meters away. there you see that both pots look very similar. The lower image was made from about 60 meters from the kettles. there you see that the mesh has already transited to LOD0. The Sculpty does that when i am about 70 meters away from the kettles.:

I think, this is a good starting place: http://blog.machinimatrix.org/kettle-quest That is an article about comparison between Sculpties and Meshes and i hope it explains a bit, what is going on. Meanwhile i have released part I of the article as Video tutorial (can be found on the page above). I am in the middle of Part II of the tutorial series where i explain how to make better LODS. I believe that the video will add a few tricks anf hints to what you already can find in the article.

Failed Inventor wrote: so how does more polygons and more prims... equal less strain and cost, verses a 1 prim slightly lower model, with nearly the same exact structure and polycount? Getting lower Prim Equivalents when you create your mesh with multiple parts is expected behaviour. As far as i understand the main reason is that smaller Objects show their lower LOD's earlier and thus they help to reduce the bandwidth. This is one of the major issues which raise when you try to make big objects (buildings). The only way to keep the Prim Equivalent small is by cutting large pieces into smaller elements. Indeed my own experiments show that for small objects you get a large discount when you link them into one Linkset. But for larger Objects for some reason linking them increases the Prim Equivalent instead of reducing it. And i hope this is a bug. Failed Inventor wrote: Also I am having trouble with manual LOD's... when mesh displays anything lower than highest lvl of LOD it breaks the texture. How do I do a single texture which works with all lvls of manually made and optimized LOD? Two easy ways to get reasonable LOD meshes which preserve the UV-Map: by construction: You start with the LOW level of detail mesh and model your base object. UV-Unwrap the model now. You subdivide your mesh once to create the MEDIUM level of detail mesh. With Blender-2.4 you can use Multires or subsurf. With Blender-2.5 you can use Subsurf or a combination of Multires plus Subsurf (i am currently working on a tutorial for that) Similar tools should be available for other 3D tools. Subdivision should NOT change the UV-layout, so this ensures that you can use the same texture for both models. You subdivide again to get the HIGH level of detail. Of course you can reshape your model and make use of the extra vertices. Just take care to not change the overall shape, otherwise you can see LOD jumping. Use a retopo tool for creating the LOWEST level of detail. In Blender you can try to remove edge loops fro example, or use the decimate modifier (blender 2.5) or the poly reducer (Blender-2.4). Or simply use the generated LOD from the SL Importer. By dedcuction: You start with the HIGH level of detail model, do all your work there and create the UV map. Now remove edge loops by hand, or use a retopo tool or poly reducer (as described above) for the other LOD meshes. In the next few days i hope to release Part 2 of the "Kettle Quest series". There i briefly show the process workflow.

i heard a rumour that you can force a sync of your inventory when you change your Second Life Password. I have not tried it but someone mentioned it a few weeks ago in one of the Mesh meetings.

I always thought that linking objects together into a linkset should be encouraged and not penalized. So we are now encouraged to only make linksets of small items ? It starts getting too much for me: I take a bowling pin (2.23 meters high)... PE=2 (selection streaming cost tells me 1.9) I create 4 Bowling pins (same height) ... PE=8 (selection streaming cost of the unlinked set: 7.6) I link the 4 pins to a linkset ... PE=9 Make the pins smaller, until the linkset PE < 8. Now the link set gets cheaper than the 4 isolated pins. I need a break.

Assuming you provided an appropriate physics shape, did you switch the Physics trype from "Convex Hul" to "Prim" after you have rezzed your object ?

I am back on mesh since 2 months, mostly testing and reporting enhancement requests to the Jira. My initial enthusiasm about meshes has been replaced by a lot of concerns. But as always i keep saying "never give up" and eagerly wait for some positive signals (like this one, which makes me smile again: http://community.secondlife.com/t5/Mesh/Loki-s-Adventures-in-creating-a-mesh-object-for-SL/m-p/954439#M2474 ) Well, we will see...

Brian Demme wrote: ... Gaia's 3 mesh tutorials got me right to the edge of where I need to be, but now I'm stuck again and have been beating my head in over it for 2 weeks now so any help would be greatly appreciated. I just have finished to convert part I of the kettle_quest tutorial into a video tutorial. It turned out to be mostly about showing initial problems, but no solutions. But in the next part (working on it) i will show a method how to make good LOD shapes and physics shape. That will probably cover "optimizing meshes for low face count" and "getting optimal LOD behaviour". I believe that pat III will eventually deal with UV-unwrapping, how to define texture faces and how to play with the texture options. But it may take me a couple of days to get there... I still plan to remake the Lighthouse tutorials later, but as long as LL is constantly changing everything it is no fun to do tutorials. Yesterday Nyx told me that they will release a new importer in the next days, so i already know that i have to remake parts of the kettle_quest Part I video.

But from the formula in the wiki, we can derive: Total Server Weight (static objects) = num_prims/2 + 1.0 Total Server Weight (dynamic objects) = num_prims + 2 See also http://community.secondlife.com/t5/Mesh/Understanding-Mesh-Server-Weight/td-p/950355 So in my above example it looks like the prims are all counted as dynamic objects. When they where counted as static objects, the total Server Weight should be 6 for 10 objects and not 12 as i see... BTW i just tested what happens when i actually add a script. And the weight does not change. So i getting more confident that something is wrong...

This matches with my recent experiences. As long as you only look at 1 Prim mesh Objects and compare these with one Prim Sculpties or regular prims, mesh looses. But as soon as you compare a mesh with a multipart build, mesh gets much better. And as far as i can see, it seems possible to make cost effective low prim meshes which can replace high cost regular builds. We only have to say good bye to 1 prim meshes as replacement for one prim Sculpties. What still makes me real trouble are all issues with meshes linked to regular prims. When they continue to end up in unexpected high costs for the link sets, then this may(will) become a big source of frustration.

Tiberious Neruda asked yesterday why his Bowling pin would not go below 6 PE. So i thought, maybe i can benefit from my recent experiments with the new Prim Equivalence calculation system... So here is my Bowling pin: Physics Weight0.1Streaming Cost0.8Prim Equivalent2OK, thats is good so far. The PE=2 is because of the Server weight (1 + 1*0.5 = 1.5 -> rounded to 2.0). Now i want to have 10 of these pins standing together: Physics Weight1.0 (as expected)Streaming Cost7.7 (close to expected, probably due to rounding)Prim Equivalent20 (10 times 2.0)Still as expected. Now i do the "link us and save weight..." magic: Physics Weight1.0Streaming Cost7.7Prim Equivalent12 (this is not what i expect)Well, this is not what i expected. Why ? Because: The PE is the maximum of (Streaming cost, Physics cost, Server Weight).Since SC= 7.7 and PW=1, Server Weight is what makes the PE go up, so SW=12But... The pins do not contain scripts. And they are not made physical objects. Hence the SW should be 1 + 10* 0.5 and that is 6. Hence i expect to see a PE of 8 here (because of the streaming cost = 7.7 becomes the highest of the three numbers). Am i wrong or is this a bug ?

Hi. I have started a video series about moving from Sculpties to Mesh. The goal is to show the differences between Sculpties and Mesh, where we can optimize meshes, how we can make better use of the resources, and lots more. I take the Kettle Quest article on my Blog as a guideline for the movie series. Part I ("departurte") is ready for review. Any comments, corrections, further suggestions are more than welcome: Thanks to everybody who gave me hints and insights to Meshes. And thanks to Vivienne Daguerre, who gave me the initial motivation to make this tutorial series.

I experimented with Server Weight in order to understand the consequences of its introduction and in order to understand how to work with it. From the experiments i only can tell that i do not understand why it was invented. From what i learned so far, you can compensate the server weight by simply reorganising your link sets. And this looks to me like that could be done under the covers, within the server, without us (the users of the system) having to deal with it. If Server Weight would be removed, we instantly would get a much easier to understand system with resource costs and physics costs only (and even that is not simple, only simpler than with Server Weight on top of it). And it would again be possible to get mesh objects which cost just 1 Prim.

Hi. Nyx has asked to do another review of the GUI. I started a Jira and added a few propositions for enhancements. Please feel free to add or discuss in that Jira: https://jira.secondlife.com/browse/CTS-664

I have seen this behaviour to. It seems to happen when the physics mesh has less texture faces than the visible mesh or the visible mesh has more independent parts ("sub" objects in the mesh) than the physical mesh (probably the latter). The weird scaling of the physics mesh may be because it scales according to one of the parts of the visible mesh and not according to the entire visible shape. I am not sure if this is a bug or a feature. Mesh is becoming rather complicated these days ;-( [edit]: I remembered that i created a jira about this: https://jira.secondlife.com/browse/CTS-615 and it was acknowledged as a bug (but still unresolved) Maybe it makes sense if you add a comment ?

heh, here is a strategy: Assume you had n scripted objects and m non scripted objects. neglect for a moment streaming costs and physics costs and focus only on server weight: Total Server Weight = n/2 + 1 + m + 2 Total Server Weight = n/2 +m + 3 The break even (number of prims == total PE) is then where: n+m >= n/2 + m + 3 n >= n/2 + 3 n/2 >= 3 n >= 6 this means on the contrary: Wherever you can bundle 6 low resource meshes, you can get one set of scripted objects for "free". So all we have to do, is to bundle our objects in a clever way. Then we can get rid of server weight costs. Well ok, this is super simplistic and it works only when the scripted objects actually CAN be bundled into one linkset. But will the server run better, faster, more stable if such a bundling is done ?

Actually i have put almost a week into my kettle tutorial to show how to convert a 2 sculpty object into a one-mesh object with a disappointing PE=2 and a lot of extra work (to be fair, i got better LOD and i got 4 texture faces, so i got "more for my money" ... somehow) But my point here is to understand the impact of the server weight (isolated from the other weights). As far as i understand the issues you mention, they are related to physics costs and streaming costs. But server weight at its own seems to be much less painfull as commonly believed. You only have to look at it from the right view point, then it becomes less freightening. Well, i still do not understand why the smallest PE is 2 and not 1, but Runitai has mentioned a few arguments which may make this extra fee understandable. Although i honestly did not (yet) understand what makes a 2 prim object more efficient than 2 one Prim objects.