Mesh houses

[Apple Pumpkins]

Yes / no? I've been reading the forums about this a lot and have many different bits from several sources.

I've gathered that the physics can be a problem and can bump the PE up, so is it recommended to use invisible prims to create physics, would this be less 'cost effective' on the PE?

Also, is it ideal to even mesh an entire house? Or would it be better to prim the basic house and use mesh components (stairs, columns, pillars, etc).

Doors/Windows with glass. I've asked about alpha sorting before and came to a conclusion on that, but would it affect doors and windows even on a seperate component of a mesh, if they're all part of the same mesh? For example I have the windows as their own mesh 'component' within the larger single mesh object, would I be able to see say the back windows from the front of the house with alpha sorting. - Is it better to prim out each window and use a flat alpha texture as was previously done?

What is a reasonable prim number for an average house? I imagine there's 'too much' in terms of how many prims a house should have but I'm unsure.

Thanks in advance. :) :)

[Ishtara Rothschild]

I imagine that texturing could be a problem if you create a large part of a house as a single mesh, since the texture size is limited to 1024 x 1024 pixels. If you don't want things to look blurry up close, it's best to build it out of smaller parts that can be individually textured.

Windows would have to be made from single prims anyway if you want to script them to open or change in opacity. And you can probably lower the overall prim count if you use megaprims instead of mesh for things like floors.

[Maeve Balfour]

I am currently experimenting with creating structures in mesh, purely for my personal use on my home plot... I'll share a few things I have learned along the way to date.

Originally I was trying to create an entire structure as a single mesh object. Although feasible, as Ishtara mentioned, it resulted in textures being spread across wide areas, resulting in blurriness / pixelation. Sometimes this is acceptable, in areas where AVs wont see them up close... say, high ceilings, rooftops, background details etc. But for up close, they don't look so good - although if you work with materials and use tiling textures, large areas could work okay. So now I break my structures down into chunks that I know will handle textures without too much overstretch, especially when combined with material mapping and overlapping UVs (as I talked about in the alpha glitching thread).

For things like windows with alphas etc... I imagine you could make them in a separate mesh, and then fit it into your main structure. Remember that you can have a number of windows, spread throughout your house (within the 64 metre limit) all still in the same mesh object. Combine this with materials, and you could save on prims there if you get creative. The alpha issues probably wont be a problem, especially if they are located in different rooms etc. However, if they have working parts (open/close etc), as Ishtara said, they would be better as invidiviual meshes or prims.

Regarding PE cost... I learned something surprising last weekend. I have been building room interior prefabs (individual rooms, corridors etc which I intend on using in a modular fashion in my builds). I had been having issues using my optimised mesh physics hulls (which I include in the uploader for SL), with them not being acurately reflected - I would be walking on air, so to speak, with the floor of the physics hull seeming to be sitting out of alignment with the actual visible mesh. So as an experimental workaround, I uploaded the optimised physics hull mesh as a mesh in its own right, aligned it with the phantom main mesh, and used that combined with a full alpha texture to hide it, and it fixed the issue. THE BIG SURPRISE WAS THE SAVINGS IN PE. The original prefab, with optimised physics included, was about 10PE. When the prefab mesh was rezzed as a phantom, with a basic 3-point triangle for its physics hull (all meshes need a physics shape, regardless of if they are phantom or not), the prefab phantom was 1PE - a 9PE saving! The mesh physics hull came in at roughly 0.5PE. So combined, the two meshes are about 1.5PE, compared to the original 10PE prefab with physics included. So things like this can shave a vast amount of PE, plus the bonuse of easier texturing combined with materials.

Also, for things like interiors, you can be savage with the LODs, because most of the time they will only be visible when you are standing inside them or close by - especially if by design you ensure they are occluded from long view by exterior walls etc. With my prefab experiments, each room is sitting between about 0.5PE to 2PE including relevant physics hulls, with reasonable amount of detailing and efficient UV and material work.

So yah, depending on how you plan it, meshes are definitely viable.

However, don't disregard prims - they will always be a valuable tool for builders. :matte-motes-smile:

[Steph Sidek]

Just an additional question, along similar lines. I've started playing with meshes for building, and have noticed, maybe it's the way I've imported it, that a mesh with holes, and different heights, seems to import with a bounding box that you walk on, rather than falling through the holes, and going up/down the steps.

Any advice as to what I'm doing wrong...

I've not started on texturing yet, just want get the meshes right before starting in on that trauma...

[i use 3ds Max 2010 to build, and straight export to collada from it, z-up. I build using 1m=1 unit. So that what I build comes in at the right sl scale ]

[arton Rotaru]

The default Physics Shape type of a mesh is "Convex Hull". Change the Physics Shape type under the Features tab in the the Edit window, to Physics Shape Type "Prim". To get the "Prim" type option you have to assign a physics shape under the "Physics" tab in the upload window.

[Steph Sidek]

---

arton Rotaru wrote:

The default Physics Shape type of a mesh is "Convex Hull". Change the Physics Shape type under the Features tab in the the Edit window, to Physics Shape Type "Prim". To get the "Prim" type option you have to assign a physics shape under the "Physics" tab in the upload window.

---

Doh, I knew it was something relatively obvious, thanks for the reply arton.

[Indigo Mertel]

Very interesting post and great tips, Maeve. Thanks for sharing.

It would be awesome if tips shared in this thread were added to the mesh wiki.

[arton Rotaru]

---

Maeve Balfour wrote:

Regarding PE cost..

. I learned something surprising last weekend. I have been building room interior prefabs (individual rooms, corridors etc which I intend on using in a modular fashion in my builds). I had been having issues using my optimised mesh physics hulls (which I include in the uploader for SL), with them not being acurately reflected - I would be walking on air, so to speak, with the floor of the physics hull seeming to be sitting out of alignment with the actual visible mesh. So as an experimental workaround, I uploaded the optimised physics hull mesh as a mesh in its own right, aligned it with the phantom main mesh, and used that combined with a full alpha texture to hide it, and it fixed the issue.

THE BIG SURPRISE WAS THE SAVINGS IN PE

. The original prefab, with optimised physics included, was about 10PE. When the prefab mesh was rezzed as a phantom, with a basic 3-point triangle for its physics hull (all meshes need a physics shape, regardless of if they are phantom or not), the prefab phantom was 1PE - a 9PE saving! The mesh physics hull came in at roughly 0.5PE. So combined, the two meshes are about 1.5PE, compared to the original 10PE prefab with physics included. So things like this can shave a vast amount of PE, plus the bonuse of easier texturing combined with materials.

---

I still don't get what's going on here. If you upload your main mesh with your physics mesh it's 10 PE. If you upload the main mesh with a single triangle for physics, it's 1 PE. That implies that the 10 PE comes from the prim type physics shape. So, now you upload that physics mesh as a normal mesh, but to have it collide properly, this requires a physics mesh as well. So what did you use as the physics shape for this second mesh? Assuming you used the same mesh for physics, as the high LOD, and the main mesh was 10 from physics, this second mesh should be 10 from physics, too.

[Maeve Balfour]

Arton: I'm not totally sure why the PE changes are so drastic (although not complaining!)

For the physics hull that I uploaded as a separate mesh, I actually used its mesh for its physics shape as well (was only about 34 triangles). So both are the same. This fake physics hull mesh came in at around 0.5PE (two of them linked equalled 1PE).

Not sure if its how the uploader combines things that results in the big difference, compared with the original mesh prefab. Mayhaps because I was extremely savage on the LODs for this separated physics mesh that it is so small in PE (since it is intended to be used with a full alpha, LOD quality is irrelevant for it).

Quite possibly this might result in the separated physics hull being so small (combination of low triangle count and extremely low LODs. All I can say is that I was very surprised at the PE difference, although definitely happy to take advantage of it.

[arton Rotaru]

I may try this at home, to see what happens :smileyhappy:. However, good you saved some prims!

[arton Rotaru]

Maeve, is it possible that you had decomposed (Analyzed) the physics mesh on your main mesh and on the second mesh upload just assigned the phyics mesh as a triangle mesh, without decomposing it? When I upload the main walls of my house with decomposing the physics mesh, I get 4.2 physics wheight with Physics Shape Type set to Prim. When I upload the same  mesh (same LODs) without decomposing the physics mesh, I get a physics weight of 0.5 when set to shape type Prim. Maybe that's the reason why you see such different numbers? I also uploaded only the physics mesh as a visible mesh (LODs one triangle only) with and without decomposing the physics (same mesh), and got the same results. Physics Weight: 4.2 (Analyzed), Physics Wheight: 0.5 (not Analyzed) on type Prim. The Convex Hull Physics Weight was the same for all meshes. 1.2.

[Drongle McMahon]

Arton. Your suggested explanation makes sense to me. I have found lower physics cost for triangle-based than for decomposed physics shapes for several building-scale shapes made up only of large triangles.

[Maeve Balfour]

Arton... Hmm, regarding your question about decomposing... it's quite possible I did what you suggested, unintentionally (it was in the early hours of the morning when I was experimenting, so I was probably half awake at the time, and could have done something like that without realising it)... If so, that would skew the results, so in hindsight may not be a valid 'scientific' analysis.

Also, I must say that I am still figuring out parts of the uploader, so I might be overlooking important steps in the procedures. I might be making things heavier in PE than they should be due to this, and that my "discovery" might actually be due to me accidentally doing something that I should already be doing in the uploader window LOL.

Ah, gotta love it!  :smileyvery-happy:

.....

Getting a bit off the track of the original OP, but I'll ask anyway since this subject has taken off:

I have found that by using a separate mesh for fake physics, I have effectively eliminated my problem of my optimised physics hulls not accurately matching my main mesh (the "walking on air" effect). Now, for a large mesh structure, where you really only need to model physics for anticipated floor traffic (assuming people won't be flying around etc), I would think you would only need a physics hull for floors, walls to about AV height, gaps for doorways, ramps etc.... in other words, you could probably skip modeling physics planes for high areas where AV traffic is not likely. So... if a separate fake mesh for physics was in this way smaller (similar ground area, but nowhere near as high as the much taller "real" phantom mesh), and it having a smaller bounding box area due to this... would this help result in a lower PE? (Compared to a combined physics hull which I assume would need to have the same bounding box dimensions as the target mesh). Smaller meshes as far as I know generally have a reduced PE compared to larger ones (ie: identical in triangle count etc)... so my theory is that a smaller "fake" physics mesh might be a PE saver in this regard when compared to a physics hull of larger dimensions, especially if you can eliminate a lot of triangles from the higher areas.

Please feel free to shoot me down if I am incorrect with my concept LOL :smileyvery-happy:

.....

And back to the OP topic... yah, as I mentioned originally, meshes have a lot of potential for houses, especially if you build in ways to make effective use of PE.

:matte-motes-smile:

[Maeve Balfour]

Indigo: Hmm... thank you for your suggestion. I had never considered adding to the Wiki before, since I am still much a learner with mesh. However, if you think my tips are valid, sure, I'd be happy to submit them sometime (although my separated physics hull idea is "under review" currently! LOL).

Do you have a link for the relevant Wiki page? (I am so sloppy with my browsing habits - I keep forgetting to bookmark stuff!)

:matte-motes-smile:

[Drongle McMahon]

"if a separate fake mesh for physics was in this way smaller (similar ground area, but nowhere near as high as the much taller "real" phantom mesh), and it having a smaller bounding box area due to this... would this help result in a lower PE? "

Short answer, no. The effect of ovearall size on physics weight depends on (a) whether the shape is triangle-based or hull-based (decomposed/analysed), and (b) what the setting of the phjyysics shgape type is (on the features tab).

Triangle-based physics shapes actually become more expensive as the triangles get smaller. So the larger your triangles are, the lower the physics weight. That is the opposite of the effect you expect. This is why triangle-based shapes are good for large buildings with flat walls etc. Of course the weight does increase with the number of triangles, and that means they can become expensive very rapidly if you want curves because you get more triangles and they are smoother. This applies equally to hull-based shapes.

Either using the Convex Hull physics shape type, or using a decomposed shape, which is a collection of convex hulls, the physics weight does not change at all with size. So making it smaller makes no difference. However, what does matter is the number of hulls plus the total number of surface points in them. This is often overlooked, especially as curved meshes have large numbers of surface points and increase the weight rapidly.

For example, if you have a conical roof, including that in the physics shape would make it expensive, while replacing it with a square pyramid, or better, a flat plane, will greatly reduce the physics weight.  Removing the roof altogether will save most, but after that lowering the walls will make no difference. Personally, I don't like people to be able to fly through the roof, so I would always have at least a flat plane. Then it depends on what you want to happen when people stand on it.

For buildings the thing to remember, for download and physics weights is: Flat = Cheap, Curved = Expensive.

Details of the physics cost calculation are described here in the wiki.

[Maeve Balfour]

Thanks for the info, Drongle, it helps me get my head a bit more around the complex ways physics works with PE. Generally, I deliberately keep my physics hulls as simplified as possible to reduce their triangle counts, so I am on the right track there... and I avoid curved surfaces for the same reason. Your mention of smaller triangles in physics being more costly is a good reminder for me (I have read that before, but I forgot that important factor, so thank you).

Lots of handy things for me to digest :matte-motes-smile:

For personal usage, I will still probably use my separated physics hulls (with alphas) to get better aligned physics in my builds... or at least until I figure out how to get consistent results which avoid the "floating on air" issue I have mentioned previously. Regardless... I am happy with my own PE results, so in the end that's what matters for me.

:matte-motes-smile:

[Nalates Urriah]

I suggest anywhere in the Mesh Section... look through it and see if there is a place you think your tips would fit. If not, just make a new page. Plus you can always ask in the Discussion or Talk pages. Also you can contact Jeremy Linden and ask where he would suggest. He handles SL doumentation.

If you do make a new page, include the right categories.