Mesh land impact

[Max Pitre]

I can get some decently low prim meshes but I am seeing whole bed room sets as only one prim. Bed, head and foot boards, end table with lamp and carpet all as one prim. The detail seems to be decent and the objects seem to have decent LODs.

 What is the secret with this? I use Cinema4D for my meshes.

[Rolig Loon]

Try

http://www.lokieliot.com/blog/?p=1193

https://community.secondlife.com/t5/Mesh/Prims-Prim-Equivalent-Land-Impact-a-too-long-guide/td-p/1293579

and make textures do as much of the work providing detail as you can.

[Theresa Tennyson]

---

Max Pitre wrote:

I can get some decently low prim meshes but I am seeing whole bed room sets as only one prim. Bed, head and foot boards, end table with lamp and carpet all as one prim. The detail seems to be decent and the objects seem to have decent LODs.

 What is the secret with this? I use Cinema4D for my meshes.

---

It may be that those aren't "true" meshes but sculpted prims, which were the common way of making complex low-prim objects before the Viewer 3-technology mesh viewers (and are still usable today). They used special textures to manipulate the vertex locations of a standard SL primitive, and with the prim accounting of the time each one was only considered a single prim for land impact purposes even though they could be very complex. When Second Life started allowing meshes from outside 3D modeling programs they introduced a different system for assigning objects "land impact" which was a more accurate way of describing the actual load an object put on the system, but sculpted prims could still use the old accounting so existing content wouldn't suddenly balloon in land impact.

[Max Pitre]

Thanks for the info. Are these threads still viable after 5+ years? Also, none of these threads seem to explain how a whole detailed bedroom set can be one LI, at least to me that is. Maybe all it is is a very clever and very experienced ( and patient ) 3D modeler and texture mapper.

[Christhiana]

I looked up the bed on MP and it is indeed a sculpted prim and not mesh. They make use of sliced sculpted prims. In the video link below it's shown how this works. The guy uses an older version of blender but it's to show how it works in general.

It takes a lot of skill to create something of this quality out of a single prim.

[Max Pitre]

wow, thats crazy. I think I'll stick with my C4D meshs and figure how to reduce LI instead. I never ever learned Blender and dont want to start now.

 

Thanks

[Chic Aeon]

I just wanted to mention that it IS possible to do what you are saying with mesh. One has to think simplistically of course and keep the poly count down. I just got a goregous bed with five pillows, comforter and many pieced wooden bedstead with lots of details, shadow prim (could have been a rug) and WITH animations  that is two LI (I would mention the designer but I don't think we are supposed to).

 

IF someone linked a whole set together and kept the poly count down it is doable. The piece of furniture  could actually be 1.499 and still be counted as I LI by the current method.

 

In order to do this the mesh needs to be simple to start with and likely have custom lODS (or it may be simple enough not to need those - who knows).  

I have actually made a couch (six cushions for three sitters) chair and table all under 1 LI (linked). I certainly could have added a simple rug (a plane) if I had wanted.  Not everyone is going to be thrilled with the inability to move furniture around though . So there are trade offs. 

 

Also you can only have 8 materials in one object so there is that limitation but it isn't a big one.

 

I work in Blender so can't speak to your program .

 

 

[ChinRey]

 

 

If you want to create really efficient mesh, the first thing you need to know are the three weights, server weight, physics weight and download weight. This is because they are calculated very differently and it's the highest of those that counts as the LI. In other words, to reduce the LI, you need to reduce the highest of those three weights:

• Server weight is supposed to be an estimate of how much work it is for the servers to retrieve the object from the database. For a single unlinked and unscripted object it is always 0.5. For scripted objects, add something between 0.2 and 0.3 for each active script. For a linkset it is 0.5 for each part plus whatever any scripts may scripts add.
• Physics weight is an estimate how much work it is for the server to calculate the physics of the object.
• Download weight is an estimate how much data needs to be transferred. The basis here is the compressed file size for each of the four LOD models. The four numbers are weighed differently depending on how often each LOD model is expected to be displayed and that again depends on the switch points. Usually the lower LOD models account for more than the higher but there are lots of exceptions to this rule.
     The switch points are the theoretical distances where the object switches between the various LOD models. It depends on the object's size and also the RenderVolLODFactor you have set your viewer to. The formula for the first switch point is: √(x2+y2+z2)*L/0.48 (x, y and z are the dimensions along the three axises, L is the RenderVolLODFactor.) The second switch point is four times as far and the third eight times as far away as the first.

Reducing the server weight is fairly easy: just combine as much as possible into a single mesh and avoid unnecessary scripts. However, big, complex meshes tend to have considerably higher download weight than multiple simpler ones. It's essential to maintain a balance here. The more you combine into one mesh, the lower the server weight but the higher the download weight.

Reducing the physics weight can be quite tricky sometimes. Fortunately, nine times out of ten all you really need for physics is a simple cube or even a single triangle which give a physics weight of 0.36 and 0.2 respectively. Also, if you can set the physics shape to none for a part of the linkset after uploading, the physics weight for that part drops down all the way to zero.
   I really can't go into detail about physics shapes here, that's way too big a topic for a single post and there are already several entire threads about it on this forum. But it is important to realize that whatever physics shape you upload will always be used just as a template for the uploader to generate the actual HAVOK physics shape. That means optimizing the physics model can be quite counter-intuitive at times. Also, be aware that analyzed and unanalyzed physics models are treated very differently (both have their advantages and disadvantages) and that the poly count doesn't matter that much. It's not at all unusual that you can reduce the physics weight by adding polys to your physics model.

Download weight is usually the factor that determines the LI. You can lower that by reducing the amount of data, that is the number of vertices and polys, or by making the file easier for the compression algorithm to handle. There are lots of tricks and techniques here of course and in the end it's all down to the mesh maker's skill, experience and judgement. Some points to keep in mind though:

• As Rolig already suggested, use textures for all they're worth. It's amazing how much detail and quality a good texture job can add to even the simplest model. However keep in mind that textures equals lag. Cover everything you have with high resolution custom baked textures and you'll end up in lag hell!
• Pay attention to the switch points and reduce each model to the details they actually need to have at those distances, nothing more and nothing less. Calculating the switch points can save you a lot of work but if you think that's too tricky, a bit of testing on the beta grid will do just as well. (You want to do a reality check on beta before the final upload even if you do the math.) Oh, and calculate and test with RenderVolLODFactor set to 1. We're talking serious mesh here after all, not the bodged up LOD disasters we see far too often in SL.
• Never use LOD models generated by the uploader. They are just way too crude to be efficient.
• Try to avoid combining big and small polys in a single mesh unless you know what you're doing.
• Consider splitting a complex mesh into several simpler ones. That will increase the server weight of course but usually also reduce the download weight. A good balance between those two is essential for truly efficient mesh.

[ChinRey]

---

Max Pitre wrote:

Are these threads still viable after 5+ years?

---

Rolig's links surprised me quite a lot too. Yes, they are a bit outdated but not nearly as much as you would expect for posts that old and there are stil a lot of good info in them.