DanielRavenNest Noe

A hollow octagon should be nowhere near 49 Prim Equiv (PE). Check the original model to see how many triangles make up each side of the octagon, it should be two, and the total triangle count about 64: two for each side - outside, inside, top and bottom times 8 sides. What drives up the PE cost of a mesh are mainly: - Triangle count, especially for the lower Levels of Detail and the physics shape - Size, if over about 4 meters it gets an increasing penalty because it is seen at higher detail for longer distances on the region. (if you don't know what levels of detail are - they are simplified versions of the model seen at longer distances. Textures do that too, a less detailed version is loaded when you are far away and it does not take up much space on your monitor)

Any kind of map adds information mapped to the surface of an object. In Second Life we just use the "diffuse color" map to give us the color at each point of the surface. Other graphics engines use more maps to give a more realistic result. There are over a dozen types that I know of. Take a look at this brick wall from another graphics engine. The wall is actually a flat surface in the 3D model. But it has a grayscale (single color ranging from black to white) that provides height information. That's called a "Displacement Map" because it actually moves the surface of the rendered bricks. Bump and Normal maps only affect the shading relative to a light source, but don't actually move the geometry. So they give the effect of a bumpy surface, but weaker than displacement when seen at an angle like the photo. Brick Displacement Map The tradeoff with any extra maps is that it takes a graphics card more time to process when you have more of them. But they do improve how things look, sometimes dramatically. Since graphics hardware only gets faster with time, I think it is just a question of when SL will add them, not if they will add them.

Medhue Simoni wrote: Yes, but you can only import to a max biped, at least in max 8, then you have to go thru a number of other conversion methods to get it to work on any other skeleton. Plus, there is no export to bvh. Max 8 is obsolete at this point. Current version is 14 (labeled as 2012). Most plugins these days dont go back before Max 2009 (v. 11). One thing that has changed a lot in the last few years (and is still in progress) is that most 3D programs have evolved a lot better import and export to other programs.

Make custom Level of Detail models for upload, it will look better than the auto-LOD. The physics shape should be just one triangle for each step and a hexagonal column for the outside, with cutouts where you start and end. That will keep the prim cost low. The SL mesh cost favors very optimized versions for the lower LOD and physics

The Magnum RC channel is supposed to be up to 10% of the main grid. Additionally some private regions have been added, and the Mesh Sandbox regions. As far as I know, if there are no serious bugs found, they will roll out to the rest of the grid and release a Mesh-enabled 3.0 viewer as the main viewer download by the end of this month.

Hi Sommer, I sent you a fixed copy of the dress on Aditi, and a download link to the 3ds Max and dae files privately by IM. It can use tweaks in geometry and vertex weights, but it's wearable. What was messed up was the scale on both the bone set and dress. They were not the right size (about 1.8 meters tall) nor standing on the 0,0,0 location, and different sizes from each other. In Max I would use the "Reset Transforms" utility to normalize scale to 100% and rotations to 0 degrees. I dont use Maya myself, but there should be a similar function there. Make sure you do the following steps or their equivalent (its what has worked for me at least in Max) * System units should be meters or centimeters * Item positioned with feet at the origin point (0,0,0), facing +Y direction, pivot located at 0,0,0 (not object center) * Item sized to match bones, and both around human size of 1.8m tall head to toe * Normalize all scaling and rotations to 100% and matching system XYZ axes at 0 rotation angles * Collapse editing stack to leave just a pure mesh or editable polygon object, remove anything besides your model (like your clothing in this case), and bone set * Apply a UV map and material so it's not invisible or untexturable. * Rig (attach bone list and vertex weights) as the last step.

If it has an equivalence of 800 and its not too large (a few meters in size), it likely has half a million triangles in the model, so it's no wonder you are crashing. Give us the specs from your PC (SL viewer > Help Menu > About second life, if you don't know - the second group of lines is your PC, the first group is the server), and the geometry data of your model (bounding box size, and number of triangles/polygons). If any part of your model has more than 64K vertexes, it will crash the viewer.

It's built into Max, at least for 2011 version: Importing HTR/BVH Files (seach for "Import BVH" in their help page - Top left Search button)

Does it show the number of triangles? Did you give it a UV map? What happens when you upload it anyway (assuming the upload button allows it)

Try turning off HTTP texture fetch, to force delivery by the other route. I know that the HTTP has caused problems for some people. Advanced menu > Debug Settins > UseHTTPInventory = FALSE

If the Collada exporter works properly, the units should be the same, so 10m = 10m. The only way to be sure is export a test object like a 1 meter cube and see what size it turns out. The maximum size for import is a 64m bounding box. In other words, your model must fit within a 64x64x64m cube aligned with the XYZ axes. Note that large mesh objects are heavily penalized in prim count for size (because they can be seen at longer distances at higher LOD levels). So large objects will likely end up lower prim cost as a sculpt, with a low triangle count mesh used only for custom physics shape (collision).

There may not be, but rigging is not the way to do what you want, since that will make an object bend and stretch

I guess I was wrong about the clip distance. Zooming in on large objects cuts off part of them if you get too close, but zooming in on a small object I was able to get 2.5 cm = height of screen, therefore 0.025 mm = 1 pixel on a large screen. Camera focussed on a large object close to an avatar, and move the camera with alt-mouse & ctrl-alt-mouse will clip the avatar at some close distance if you swing the camera close enough, so it may work based on the physics shape/collision boundary. I could not find a debug setting for it, but is is clipping and limiting view distances. @Mikki - you can weight a given vertex of an object to any of up to 4 bones, but then it will stretch and bend in response. If you want a rigid object to move with two attach points, like both hands, that might require some scripting, attach to one, then take the other arm position and adjust the attach point parameters to follow it

One thing to consider for avatar attachments is the camera clip distance is normally 0.5 meters - anything closer than that cannot be seen. At that distance, one monitor pixel is 0.5 mm. Thus any geometry or texture details smaller than 0.5 mm simply will not be seen. For gently curving items like heels and shoe uppers, the triangles that make up a mesh can be considerably larger than 0.5 mm before the curved shape has noticeable flat sides, especially if you use smooth shading. Nicely detailed attachments are great, but excessively tiny details are just a waste of geometry and textures. @Mikki - you can wear non-rigged attachments the same as any prim attachments, and edit their position and size. Rigged items are for when you need it to bend, like a sleeve, and follow more than one avatar bone.

There is no need to save a new version every time you move a vertex. Mesh are displayed from a file on your PC. Editing a vertex simply changes a few numbers in the file (the position of that vertex). Your PC can easily hold the modified copy of the asset and use that for display while you edit, and then only send the changes when you finish the editing session. So long as the only copy is sitting on a region, the region can hold the changes. So that object is base mesh asset + some set of changes stored local on the region server. When you take a copy to inventory, that is when you generate a new asset, same as any other edited object.

Tne Maxon Cinema 4D website says the current version can export to "Collada .dae" file format. That is what the SL model upload needs. If you have an older version you will have to check to see if that is a supported format for export. I've never used that progam, so I can't help you with the specific details, but you want to create the geometry as a polygon mesh with triangles or quads, and apply no more than 8 UV mapped texture IDs. Scaling should be metric, either cm or meters.

Nonetheless I opened a JIRA issue requesting a vertex editor - the simplest possible way to modify a mesh: vwr 23202 It's been ignored since last fall, so I don't think any more full featured editor will happen within SL.

Hi Sommer, welcome. There is a plugin by Wiz Daxter for SL avatar to Maya: Wiz Daxter He also has a link to a an SL marketplace version that has more features. Please note that software version matters. It may not work for 2012. All you need is a copy of the skeleton, though, and that might be extracted from the file, or you could load it in an older version of Maya and then save it, and reload it in 2012. Note also that there were some issues with the Autodesk Collada FBX 2012.0 exporter. The SL importer was having issues reading the resulting file. That may have been fixed, I have not tested it recently (I use 3ds Max, but I think that shares the same FBX exporter as Maya). If you are new to 3D modeling, this is par for the course. There are many 3D programs, and many file formats, and compatibility issues are very common. The best advice I have is take notes on the workflow you use, so when you find a way that works, you know how to repeat it.

I understand that, having written many wiki pages over the years. But this particular page was written by a manager for the mesh project, so I thought it reasonable to discuss changes first, rather than just jumping in and changing things.

I'm not sure if you want to get into this level of detail, but in the second page, under "Implications for Creators", here is some background why some models are "inefficient": Those high polygon models are not inefficient. They are simply designed for a different purpose. People doing architectural renderings, or offline animation can afford to take minutes per frame. Therefore the models they use can be more detailed. A printed full page magazine photo or ad consumes 2550x3300 pixels, several times the detail of the average computer monitor. So if you are using 3D models to generate the image, you need more detail in the source model. On the other hand, games and virtual worlds are a real time situation. Rather than minutes, you have a few tens of milliseconds to generate one "frame", and screen resolutions are lower than for the purposes mentioned above. So lower detail models are both required, for speed, and acceptable from a resolution standpoint. You might consider adding something like this to the end of that section: "Creators should bear in mind that model geometry and texture detail that appear smaller than one pixel when viewed on a monitor are unnecessary, since they are too small to see, and merely slow down the Viewer." If it's not there already, we might add a page on optimizing mesh for SL and link from that section.

IP is short for "Intellectual Property", which includes Copyright, Trademark, Trade Secrets, and Invention and Design Patents. Those rights are all different from each other, but share being intangible and being capable of being sold and infringed upon. The latter makes it possible to go to court over Intellectual Property Rights. A given item may have multiple rights attached. For example, the ingredients in Coca Cola are a trade secret, the machines in the bottling plant may be invention patented, the shape of the bottle may be a design patent, the brand name is a trademark, and the bottle label may be copyrighted. The general way IP rights work is they allow the owner to do certain things, and everyone else may not do those things without permission. If you do those things anyway, without permission, that is called an "infringement" or "violation" of the rights the owner has. For example, copyright gives the owner exclusive right to make copies of a creative work. There are limited exceptions, but in general you may not make copies unless you are given permission to. The specific details of what rights apply go by country. Almost every country is party to treaties that respect rights granted in other countries. How strongly rights are enforced varies, and there is a dichotomy between IP rights which say you may not make copies, and computers and the internet which have making copies as a fundamental feature.

Second Life does have "Materials", just very simple ones. We have control of diffuse color and opacity, and limited control of specular sharpness and bump mapping. I think the most correct description is that we can use up to 8 UV mapped textures with 8 material settings. Each UV map can include an arbitrary subset of the triangles in the model arranged in an arbitrary way, but each LOD level needs to have the same number of maps, and at least one triangle on each map. Only the mapping is part of the model upload, which diffuse texture is applied, and the other materials settings, are selectable while online.

Vivienne Schell wrote: What is "mainstream" about "mesh"? It's used by about 90% of 3D work worldwide. The balance is things like NURBS, procedural generation, and a few experimental methods like voxels. So it is quite meainstream. It is another tool which someone can learn in Second Life and go on to use in real life, along with textures, audio, animations, and programming. Second Life is entertaining for some, for others it can be a great learning platform.

It depends. Some people logged into a low power machine may be using a text-only client, or leaving it idle just to recieve messages or as a shop modeling avatar, while they are creating or doing other tasks on their better PC. We can't tell from the statistics how many people are doing things like that vs. actually only have a low power machine with no alternatives. Conversely, does the class table properly report multi-GPU systems, or people driving two monitors with one GPU (which would slow down SL on one monitor because the GPU spends time driving the other)? What bothers me is statements from Linden staff that they are targeting mesh performance for systems below their stated System Requirements page. Either the system requirements are wrong, or they are designing mesh for PC's they don't actually support people using.

Version 3 of rock texture. Now it looks like crusty cheese bread :matte-motes-big-grin: What I did this time was bake the ambient occlusion from the high poly version onto the rock texture.