Jump to content

Senkiya Zsun

  • Posts

  • Joined

  • Last visited

Everything posted by Senkiya Zsun

  1. Not sure what software you are using but normally all faces start with a default texture which is a big gotcha. Those defaults can creep in and cause that error message. As for relying on auto-generated LODs, it's really not something a professional should be doing. The great majority of residents are not LOD aware and won't have the reflex to check a product out from different distances before buying. Not one auto-LOD I've seen when uploading stuff I've made has been good enough to use – not even near.
  2. Again, just make the lower LOD exactly as you want it and add barely visible triangles (you'll need three at the most) to make it fill the bounding box. I'm asuming the LOD isn't bigger than the main one; I can't see why it should be. If it is then add the tris to the main LOD instead. ETA: best to set RenderVolumeLODFactor to 1.0 while testing your LODs inworld (for obvious reasons, which I'm sure you already realized).
  3. Yes it's materials. It's a bug they keep trying to fix without success. Well that's not quite true: the "subset" error is normal: it's normal to require low LODs not to have materials the main LOD doesn't. But there's also the "same number of texturable faces" error which insists on lower LODs having every single material the main one uses; and that's not normal. For the latter you have to find a "throwaway" surface on a lower LOD to dispose of the unwanted materials on. With it being a lower LOD, this isn't usually a problem as you can make the offending face one that's hardly visible at a distance. If not, then you have to add a tiny triangle somewhere for each one. You're also going to discover the joys of having to make all the lower LODs the same size as the main one. But seriously, custom LODs are the way to go. I dread what SL is going to end up looking like if a lot of people use the automatically generated ones, which are approximate at best. Everyone will eventually set their RenderVolumeLODfactor to 4 and LODs will then be pointless, and all the sweat that went into them will have been for nothing.
  4. Pamela Galli wrote: Well I made a physics shape for my door and all went well so now am considering making them for the walls. I can't upload a physics shape separately and stick it in the walls of the house like I would SL prims, can I? Of course you can. Just make a very simple, solid mesh with the same overall dimensions as your wall, add your physics shape with all the holes in, then make it invisible. Have I misunderstood something? ETA: sorry just noticed Maeve says the same thing at the end of her reply.
  5. While on one hand I'm happy that it appears a single scripted door, for example, will no longer double the server cost of my entire building, which is composed of many meshes, on the other hand it is going to turn some of my "land impact 1" objects into liars. The only solution to restore the truth will then be to combine the scripts. And since I split the scripts in the first place to make them more efficent, combining them back into one script is going to defeat the intended purpose of this proposed change. I agree with their idea, but "number of scripts" is simply the wrong metric, too rough and ready.
  6. Learn how to build models first and worry about the poly count later. Because the first takes about 5 minutes and the rest takes months. Seriously.
  7. Using the limited creation I've done so far as a yardstick, I could carpet a whole sim with end-to-end high detail mesh buildings stacked three high without busting the available LI. That gives me a good idea of what's reasonable for static objects.
  8. You know, I like to be technically correct whenever I can, but sometimes when the optimum result amounts to bucking the system then that's the way to go. As I think I said, this was an exceptional case since usually I manage to make a suitable LOD4 without kludging. But yes it totally would be nice to get this issue sorted out. You should be able to manufacture a very simple LOD4 that doesn't affect the LI of the whole model without having this kind of worry. Sen
  9. Bent Melody wrote: OK, I created a cube to contain my object, yet when I attempt to use it for the physics file it only surrounds the root prim of the object (in the view window of the uploader) no matter how large I make it? ughhh You need a cube for each component. All lower LOD or physics shape elements fill the bounding box of just one highest LOD component. So if you only put one cube, it will only correspond to the first component.
  10. I think the key word here might be "welded". Generally speaking a linked set of 4 separate models will have a lower download cost than the 4 combined into a single model. So provided the server cost (usually 0.5 per model/mesh/whatever you call an indivisible element) remains lower than the download cost, you gain nothing by welding apart possibly from simpler material management ... but you stand to lose an awful lot as regards LI.
  11. Coming back to the original question, if you want a concrete example, you could fit 150 of these onto the sim: Outside: Inside. I imagine your huts are a little less complex. http://slurl.com/secondlife/Kawaii/198/216/21/ Caveat: the build has been developed since I took that pic so if you visit it now you wouldn't get more that about 120 of those on a sim. And to make one more thing clear: this is a model with every LOD visible and visually correct at the standard (very low) default viewer setting, i.e., RenderVolumeLODFactor = 1.0. Sen
  12. There are three parts to learning mesh for SL. First is making the model, which is probably the easiest part. Second is texturing it, which primarily involves UV mapping and can be pretty laborious at times. The third and for me the most time consuming is optimizing it for SL (after getting used to all the SL mesh upload quirks). There are plenty of tutorials for the first and second but the third is pretty much a case of reading through this whole forum then playing it by ear. Just a word of advice, before looking at any Blender tutorials look at as many modelling programs as you can till you find one whose interface you are comfortable with. Blender does a good job but it's not everybody's cup of tea. Good luck!
  13. Camera controls can be very useful in build mode as an alternative means of adjusting your view. I always have them on display when building. For the majority of situations the mouse+key method is of course a lot quicker and more convenient so I put the camera control widget to rest.
  14. I tried using a different exporter but the results now seem even more confusing. - when I upload without a UV map (which I achieve by deleting the map from the LOD4 model), the download cost of the'whole object is 1.764. This is actually less than last time i tried, presumably due to the randomness you talked about. The tri count is 53 and the vertex count 65. - when I upload with the old exporter (wings 3d) and a collapsed UV map, the vertex count rises to 113 and the download cost, again for the whole object, is 3.082. - when I upload with the new exporter (3dsmax) and the same collapsed UV map, the vertex count drops to 74 - BUT - the download cost rises to 3.25 :womanmad: I can show you the collada files if you tell me how to get them to you. ETA: I tried again using the Blender exporter but I got the "different number of texturable faces warning" and I don't even want to think about going that way. It did however tell me the vertex count was 113, the same as with the Wings3d exporter. Sen
  15. This is getting slightly beyond the scope of this discussion, I know, so feel free to answer on a new thread if you think it more appropriate. Can you please clarify exactly what you mean by collapsing the UV map? I tried scaling the map on both axes to 0.0, thus converging all the coordinates to the same point in the centre, but it still adds around 40 vertices to the upload and 1.2 to the LI, which might not sound a lot but is a whopping 60%+ increase. I was unable to improve matters by playing around with normals. Now if I upload it with no UV map but generate normals (68% fold), it adds about 0.6 or 30% to the LI and looks just as good as with the collapsed UV map. But it's still a penalty I'm not willing to pay. To sum up: If I use neither normals nor UV map, the LI is 1.9.but I need to cheat by using forced shading. If I use just normals the LI is 2.5 and I don't need to cheat If I use a "natural" UV map the LI is 4.2 If I use a collapsed UV map the LI is 3.1 The last three all look the same viewed from LOD 4 distance (just as a reminder, the issue is purely with the lowest LOD). It's a bit annoying that I can add 100 triangles to the main LOD with hardly any effect but adding a ridiculously small amount of detail at LOD 4 can more than double the LI. I have to add that this particular object is the only one I've had to find this kind of work around for. In any case I appreciate the guidance; thank you. Sen
  16. Senkiya Zsun


    In a similar vein, I made a wall section with slats (see below). At the highest LOD the slats are actual polygons. In every other LOD they are a flat surface with a texture and therefore the poly count is obviously slashed. I think this would work well for your racquet.
  17. I never thought about collapsing the UV map to a point; that sounds like a great idea. I'll be back with the results!
  18. It seems to me a lot of people are having trouble getting their heads round physical shapes and probably LODs too. So I thought I'd offer an illustration as my contribution to avoiding bad mesh in SL. This is a stone lantern I made, showing the finished object, four levels of detail (LOD), and the physics shape. It's designed to look correct from all distances without asking residents to change their viewer's debug settings. To ensure this I always make the prototypes with RenderVolumeLODFactor set to 1.000. As you can see the bottom level of detail is very rudimentary but from a long distance it still looks like a pedestal lamp (and still glows). The triangle counts in each LOD are respectively 1468, 906, 406, and 53. The first two LODs are uploaded with normals generated while the lowest LOD has no normals and not even a UV map (it seems to be this level that has the most drastic effect on download cost and removing the UV map helps – I make up for it by using a strategic arrangement of light and dark materials). The physics shape is quite clearly two convex hulls (I could have made it one but I wanted people to be able to stand on the dome!). In every mesh I've made so far apart from one very complicated staircase, the determining factor for LI has been the download cost, i.e., triangles and vertices. In the example of the staircase, the physical cost only just exceeded the download cost. The server cost has never been an issue except to a very minor degree in a scripted linkset. The land impact of the single-mesh lantern without the light is roughly 1.8. Because I use a separate scripted element to give the lighting effects, the LI for the whole linked object is 2 (2 x 0.5 server cost x 2 for the script), I think this kind of efficiency or better still, without compromising appearance, is what all mesh creators should be aiming for. On a side note, every visitor to my parcel has remarked on the speed with which these mesh objects rez. Getting the triangle count down isn't trivial and you have to learn the tricks as you go along, but the reward IMO is well worth the effort. Sen ETA: I forgot to mention that it's often useful to seek an ideal trade-off between download cost and server cost. Breaking a complicated shape down will almost always save on download cost but break it down too much and you will hit the server cost wall. There is a secondary issue as well when doing this: the smaller the object, the closer it is when the LOD flips meaning your lower levels of detail will probably need more triangles to keep them looking good. :womanfrustrated:
  19. One of those too-good-to-be-true moments then, sigh: :matte-motes-confused::catlol: Update: I did a few tests and they pretty much concur everything you said. I couldn't reproduce the odd behaviour at first until I realized even a renderVolLODFactor of 1.0 wasn't making the LOD flip (it's quite a big object). In the meantime, I'd changed to Catznip to check what that viewer made of it, and the mere fact of changing viewers borked the high LOD material faces too (until I sat down LOL). Also, when I changed back to the Dev viewer, another strange thing happened, though I haven't been able to repeat it: not just the wrong texture was displayed but the rotation and offset parameters were changed to those of the "false" material too.
  20. I'm using the same viewer since December."Second Life 3.2.5 (245937) Nov 30 2011 13:43:09 (Second Life Development)" And yes, it uploads and is correct. And yes it still tells me LODS have a diff number of texturable faces and crosses them. Update: yes I managed to upload it on the main mesh too. :smileyvery-happy:
  21. Senkiya Zsun

    Is this new?

    Different number of texturable faces on LODs but upload works anyway and the materials on the uploaded object look OK at the lower LODs. :matte-motes-big-grin: Having said that I haven't tried on the main grid yet.
  22. PSP was great until Corel took it over (PSP 10 and later). I personally find the interface infinitely more user friendly than that of Photoshop. There aren't any features missing for RGB work that I know of (CYMK is another story), or certainly none that have held me back. For SL texture creation, the only problem is saving your work as a .tga file as there seems to be an alpha issue. I solve this by exporting as .psd and saving as .tga through GIMP (kludgy I know). I'm not sure you can still get PSP9 though, because PSP X sucked so bad I gave up on it. An example of its badness: you could only run it as administrator and it was hell on earth to put your files where YOU wanted. Having said that, I haven't looked at more recent versions to see if they have redeemed themselves.
  23. You're definitely polyvalent Knowl! Great show. And thanks to the people who said kind words about my work. :matte-motes-inlove: Sen
  24. What a fantastic collection so soon after mesh was introduced, though I guess we had plenty of time to prepare! I was away for most of 2011 and was delighted to see mesh up and running when I got back. :matte-motes-big-grin: So I decided to rebuild my dojo and give it the details it had always been missing along with a bit of extra realism! The "flaming sphere" roof adornment is all mesh. Those are real holes, not texture. This is me, to give you an idea of the scale. Roof brackets and individual wall slats. Interior very much inspired by the film Inception! Lower LOD (some medium, some low). This and the next view are taken with RenderVolumeLODFactor set to 1.000 At the limit of the draw distance (256m). For the physical shape, in every place it could reasonably be said to matter, what you see is what you get. If you walk around the curved roof, for example, your feet pretty much stick to the surface. In places where it doesn't matter (the roof brackets for example) the physical shape is set to none. And the land impact cost of all that? Including the interior stairs, railings, and floors, and the slab with perimeter fence, 98 prims. :matte-motes-big-grin-wink: Anyone who wants to visit, just look up Senkiya's Dodgy Dojo.
  • Create New...