DanielRavenNest Noe
-
Posts
800 -
Joined
-
Last visited
Content Type
Forums
Blogs
Knowledge Base
Posts posted by DanielRavenNest Noe
-
-
@leliel - If I understand what you are proposing, then maps besides color would still be uploaded as JPEG 2000 files (which is what everything is stored as right now). So normal maps, specular maps, etc. all are treated as simple textures are now, which means no change to the asset database system. But when used in the viewer edit window, we have some means of flagging "this is a normal map" when you apply it to an object. The set of maps a viewer can render depends on the shaders it uses, but I assume would be a superset of whatever is in the official viewer.
I can think of two ways to flag what kind of map is being used. One is a naming convention, where you have a string like "[normal]" in the inventory name to identify it as a normal map. The other is a list in the edit window where you select the map type.
-
You will have to ask Maxwell Graf, who I think wrote that page, why he wrote it that way. My comments are based on my experience making rigged clothing for the Blue Mars world, and testing things in the SL mesh project since the beta started. A working deformer will help a great deal to make clothing fit, but it cannot perform miracles. The base SL avatar deforms badly when you use extreme slider settings, and clothing influenced by those deformations will be equally messed up if you don't make a version for that end of the settings. By badly I mean textures will be stretched, and there are simply not enough polygons in the base avatar, so it becomes lumpy. As a clothing designer, you can work around those issues by making more than one version to fit different general body shapes.
-
Bonnie Brandi wrote:In the skin parameters i had open my Weight Table, i have some 1.0 and some 0.0, but i don't know how to do the correct weight.Could you please give me more information.___________________________
On the Skin modifier, click Parameters > Edit Envelopes, and under that Select > Vertices
Lower down, under Display > check Show Colored Vertices and Show Colored Faces (this will show the weights on a selected bone as colors on your model)
Select a bone, then select some vertices you want to follow that bone. So the hips area of the skirt needs to be mPelvis selected, then choose vertices around the hips.
Under Parameters > Weight properties, you have several ways to adjust the weight for that bone:
Abs. Effect. will give whatever number you type in (from 0.0 to 1.0) to all the vertexes you have selected.
The wrench tool to the left of "Weight Table" will open a floating window with some tools to set weights
You already found the weight table, its like a spreadsheet, and you can just type in values.
Paint Weights is like a brush, and the "..." next to it has settings
All of these ways change the weight values, and you can use what tool you like best. What you need to do is for every vertex on your model, have the total weights be 1.00. On some parts it will be all on one bone, but at the joints, like where the leg meets the hip, you need to blend the weights between the two bones smoothly, so it will bend nicely. This will take a while. It took me longer to do the weights than to make the model in the first place.
-
After you added the bones, you need to actually weight the model vertexes to whichever bone(s) it needs to follow. For a skit it's generally the mPelvis, and the hips.
-
Possibilities:
* Is the avatar facing the correct direction, and reset transforms so that all rotations and scaling are nulled out"
* Do the bones line up properly with the model you are making? If they are offset from the center of the limb or not the right length, it will end up distoted.
* I think you mean "Skin Wrap" modifier. After applying that, you generally have to hand tune the weights, or adjust the range settings in the modifier for how far away they affect vertices. My experience with Skin Wrap is it gives a first approximation by copying weights from another model, but that is not a substitute for hand adjusting the weights.
* Vertex weights need to add up to 100% regardless if it's one bone or several. One trick is to weight everything 100% to the pelvis before using Skin Wrap or the bone capsules, or other methods to do weighting. That way every part will at least have a weight to *something*.
When making my full body mesh avatar, I literally did 100 uploads to the beta grid while tuning weights. It was my first time, but I don't think there is a "make pretty" button on any 3D software. You need some amount of going back between SL and your 3D program adjusting things until it's right.
-
That description is simply wrong. Rigging is defining which bones a mesh object will be animated by. Without that, clothing will not bend, since there would be no data telling the viewer how to move it. The need to make multiple sizes, alpha maps, and adjust bone weights will be *reduced* but probably not eliminated. The reason is that even with a working deformer, when the avatar body is changed a lot, clothing that fit properly will need to be reworked. For example, in going from breast size 0 to 100, you likely would want to add triangles to maintain a smooth shape, and rework the UV map for texture distortion. Still, doing 3 sizes would be better than 25.
-
The SL character folder is 3.5 MB zipped, which is how it is delivered. In case you have not noticed, when you install Second Life, the program folder grows to around 100 MB on your hard drive when it's unpacked. A new and better avatar set added to the existing one might add twice that to the download, say 7 MB.
If you think 50 MB is a huge download, then you are still living in the 20th century. Netflix standard definition movies run 170 MB per hour, and my AppData/Local/SecondLife folder (where textures are cached) is 300 MB. Just idling in SL I use 9 MB/hour. If I am actually doing anything it's a lot more as textures download etc.
Some comparison 3D program sizes:
Autodesk XSI Mod Tool - 607 MB
DAZ Studio 4 - 446 MB
Vue R10 (digital nature software) - 2.59 GB
CryEngine 3.3.7 SDK - 754 MB
Autodesk 3ds Max - 3.5 GB
Crysis 2 Game (as installed) - 10.7 GB
-
Do you know about "smoothing" for mesh objects? Without it, a flat side is rendered as having the same lighting across the whole side, so the edge is noticeable. With smoothing, a side is rendered with varying lighting across a face (technically the normals are interpolated, which changes the lighting angle, and thus the color). This is something you select in your 3D program. Once in SL, it can't be changed for a given upload.
-
Look up the triangle count for your mesh version. If it's more than 120 you are doing it wrong. At high detail, it can be a square rod for each piece. 4 rectangular sides on each rod x 2 triangles per rectangle = 8 triangles per rod, and you have 15 rods. For the lower LOD levels, you can get away with a two sided alpha plane. An alpha plane is a simple rectangle (2 triangles), with a picture of your scaffold on it. You can render that in your 3D program, or photograph the high detail version in SL after uploading it. The holes between rods will need a transparent alpha channel on the texture (which is why its called an alpha plane). Since SL only renders one side of any geometry, to see it from both sides you will need two alpha planes facing in opposite directions, for a total of 4 triangles. The Physics shape can be the simplest version of all, that can be a single copy of one of the alpha planes, just a simple rectangle. If you want to be able to walk between pieces, though, you will need a more complex physics shape, just don't make it more complex than the minimum that the avatar and other objects need to collide with.
-
How you create the different LOD levels does not matter. Different 3D programs have different tools to help with it, and your personal workflow can vary. What matters is when you are far away from the object, you simply cannot see fine details, since they will be less than 1 pixel in size on the screen. So to save work for everyone's viewers, you should create a lower detail version to see at long distance. How much detail you remove is up to you, but the higher the detail you keep at lower LOD levels, the higher the upload cost and Land Impact values will be. They are set that way to encourage you to be efficient.
Some hints at making things look nice without excessive triangle counts (this applies at any LOD level):
* Use smoothing groups if your 3D program has it. That varies the shading across an object smoothly, making it appear rounded even without a lot of sides.
* Use alpha planes. These are flat objects (2 triangles) with a *picture* of the object in more detail pasted on it. A plant with 900 triangles close up can be replaced with a few alpha planes (4-6 triangles) for the lowest LOD. Those have a rendered image of the high detail version on them. At the longest view distance, you cannot tell the difference.
* All your LOD versions are forced to the same "bounding box" as the highest LOD, no matter what size they are modeled. A bounding box is a box prim just large enough to contain the whole model. Keep that in mind as you make the other LOD versions.
-
Peggy Paperdoll wrote:Had LL started with a more complex platform and insisted on creators optimizing the content before they were allowed to upload to SL then SL would certainly be less laggy and avatar weight would not matter as much. Blue Moon tried that and look were they are today......and some very smart people developed Blue Moon. Had LL started out with something different SL probably wouldn't be here today.-------------------------------------------------------------Umm, Peggy? The name is "". Blue Moon was a detective agency in the "Moonlighting" TV series, and a number of other things, but not a virtual world. As far as the optimizing content issue, we were limited to 512KB compressed data per clothing item, and 5000 triangles. That is actually not restrictive. I was able to do some very nice items within those limits. In a few cases I needed to upload it in two parts, but usually it could be done as one. By comparison, the base SL avatar is around 7000 triangles.
-
I looked for it in Phoenix 1.6.0 but could not find it, do you know exactly which menu path?
-
Jenni Darkwatch wrote:All true, if you only consider realtime rendering. The output of (consumer) GPU rendering is pretty damn crappy though. There's at least two (commercial) hardware render cards that are not triangle based - one uses raytracing (CausticOne), no idea what the other (RayCore) uses. Both beat the pants off any triangle based 3D render.
Siliconarts RayCore is also a ray tracer, but it is not a card yet. It is intellectual property available for licensing, to be incorporated into graphics hardware. As far as beating the pants off a triangle based renderer:
"RayCore® enables 3D contents to process in between 14 million to 24 million rays per second per core, the industry's fastest performance, and supports display resolutions of 960x640, 960x480, 800x480, and 300x240."
960x640 is not exactly earth-shattering resolution.
The Nvidia owned mental images "iray" rendering engine is also a ray tracer, and will run on any recent CUDA enabled Nvidia card, like my GTX-260. It takes advantage of the many parallel pipelines in the card to run pretty fast. I don't know how it compares to the RayCore on speed, but it definitely beats it on availability, since I have used it since last spring.
As far as Caustic (now owned by Imagination Technologies), they appear to have software tools available, but I don't see a raytracing chip or card available. The last info I find online is as of September, that the "Caustic Two" PCI card is forthcoming. If you have better information, please provide a link, I would be very interested in faster rendering.
-
Jenni Darkwatch wrote:Not necessarily. Today, most renderers are triangles. However, there were some oddball render engines around that could indeed handle ngons, as long as they were flat. I'd suspect some render engines even today can handle ngons too, though I don't know of any non-raytracing ones that do. The old DOS AutoCAD renderer had ngon surface rendering. I wouldn't call it realtime though.
Modern graphics card do just one thing, and do it very efficiently: shade triangles. They process triangles because that is the simplest shape with an area. By shade I mean color the pixels according to the surface texture of the model and whatever light sources are in the scene. The efficiency comes from having multiple "pipelines" all doing the same thing in parallel (shading triangles), but on different parts of the frame. My graphics card has 216 pipelines, and new high end cards can have 1024 or even more. By only doing the simplest possible shape (triangles), the silicon circuits that make up the pipeline are small, and they can cram a lot of them into one graphics chip.
The most efficient kind of triangles to process are triangle "strips". Take a long rectangle and divide it into smaller rectangles on the long axis, then divide each small rectangle into two triangles. This is a triangle strip. Each triangle going from one end to the other shares two corners (vertexes) with the previous one. So to load the next triangle, you only have to replace one vertex with a new one. The two that are shared can stay in memory. With disconnected triangles you have to load three new vertexes for each triangle. This is one reason to use quads with good edge flow in making your models. They can easily be turned into strips of triangles, which means your graphics card can process them faster.
Software renderers that use your CPU chip are basically obsolete at this time. At best even a professional workstation might have two CPUs with 6 cores each. With hyperthreading that means 24 threads running in parallel. If you have a triple GPU setup with 1024 shader pipelines each (and a Very Big power supply to run them), you would have 3072 pipelines running in parallel, so it is simply no contest.
-
3ds Max 2010 with the 2011.3 FBX plugin works best in my experience. If you use Max 2012, you need save to 2011 FBX format, then use the standalone FBX 2011.3 converter to get to DAE. The SL uploader seems to not understand the 2012 DAE file format. If you use Maya or a different year of 3ds Max than 2010 or 2012, I don't have those, so I cannot give a definite answer from experience, but I would expect Max 2011 should work with the 2011.3 plugin also.
Please take a look at the SL Wiki page for 3ds Max. If anything in there does not work as written, please post here, let me know directly, or correct it yourself. Both Autodesk products and the SL Mesh code evolve, so things can get out of date.
-
1
-
-
Sae Luan wrote:I didn't know anything about this. Can we still donate funds to the project? I'd love to help out.
Hi Sae. The fundraising is finished, and Qarl Fizz (former Qarl Linden) is now working on the viewer code. You can check the Snowstorm issue for progress: STORM-1716
At this point, the best way to help might be to provide Qarl with some test models when it gets to be time to test it out.
-
Mircea Lobo wrote:The second thing that pissed me off is that you are required to have payment information on file in order to upload meshes. So if your SL account is not connected to a bank, you cannot import meshes although you can upload as many images and sounds as you like. Where is the logic in this guys?------------------------Both the IP tutorial (in the form of a test) and the requirement for payment information are to cover Linden Lab legally when the inevitable upload of an item from a commercial game or other copyright owner with money and lawyers happens. Linden Lab can then say "we told them not to do it, see, here is the tutorial we require", and "here is the uploader's contact information, go sue them, not us."Uploading stuff you don't have rights to is already against the Terms of Service, but with this tutorial, someone cannot claim they just clicked the TOS button on first login without reading it. Your chances of getting the right answers on the tutorial without actually reading it are negligible. When viewed as a way to avoid costly lawsuits, it makes good business sense from the Lab's point of view. It's not there because of how the Lab wants to deal with *you*, it's there because big commercial IP owners are sometimes lawsuit-happy.Mircea Lobo wrote:First of all, how can meshes increase copyright problems more than other forms of creation SL has for years?------------------------Because in the case of models ripped from commercial games, Linden Lab is a direct competitor in the sense of it's another multi-player 3D environment which they make money from. The new Linden Realms areas are actual games, so that makes Second Life even more of a direct competitor to other games.Ripping textures from a site like Renderosity or Turbosquid is not on the same scale, because those IP owners don't have legal departments like Electronic Arts or Disney.
-
The uploader in the Viewer will accept up to 4 "level of detail" (LOD) models for showing your item at different distances, plus a "physics shape" model for how things collide with your item. If you do not supply a physics shape, it automatically creates one of the type "convex hull", which is like a shrink-wrapped shape around your item. Convex hulls have no holes.
For a physics shape, it should be as simple as possible, just flat planes for the floors and walls, leaving gaps for doorways. If you dont expect to fly through the roof, you don't need anything up there. Keeping the physics shape simple will help keep the the "Land Impact" (used to be prim cost) down.
The different LODs and physics shape are different models, but you can align them more easily if you build them all in one blender file, save it, then only export each part separately. It might be easier to delete all but one model, save that separately, and repeat for each LOD & physics.
You can use the same model for more than one slot in the uploader, but that may end up costing more land impact.
-
It's probably time to test my items that failed with Max 2012 and see if the importer is completely fixed. Max 2012 writes a Collada version 1.4.1 file, and previously the SL uploader could not read that correctly.
-
I don't have problems with scaling, but here is what I do: First, start a new file, and generate the SLAV skeleton. Resize it until it is around 180 cm tall, and facing and positioned correctly. Then do reset transform from the utility tab on the command panel, so all rotations are set to 0 angle, and scaling is set to 100% internally. Then import your model and do the same thing. Finally, use the skin modifier to link your model to the bones. If all scaling is reset to 100% in max, it *should* load up in SL correctly (at least it has so far for me).
If that doesn't work, contact me in-world and we can try to figure out where the problem is.
-
Hmm, Mesh Sandbox 23 worked for me. I logged in there, rezzed a mesh object. I do have a premium account.
Viewer: 3.2.3 (244722) from Viewer development build page
System: i920 quad core / Windows 7 Ultimate / Nvidia GTX260 card / 260.99 drivers
-
The uploader in the viewer will scale all your LODs to have the same "bounding box" dimension, regardless of what the actual model sizes are for the lower LODs. This is annoying, but that is the way it works. Bounding box is an imaginary box just large enough in X, Y, and Z dimensions to contain the whole model. If your lowest LOD is significantly smaller, you can either resize it, or add an invisible triangle near the edge to force the bounding box size.
-
To answer some questions:
* The default for already existing items is a good question. Right now rigged meshes *partially* deform, so either choice ends up messing up the items. The best answer is for the original creator to upload it again and distribute it as an update.
* I fully understand why Qarl and the Linden staff would want to do a simple version for the first release, but adjustments for bias and strength will help whenever people are wearing multiple mesh items (which is highly likely). Bias adjusts the distance from the bones a small amount, and also determines the rendering order. This will help with layering of items. Render later means it will show on top. Strength would adjust the deform amount as a variable. As described in the JIRA, we will have a setting of "on" or "off" for deforming, so essentially 100% deform, or 0% deform.
Strength would give you values in between, or even beyond (ie 200% strength). Again, this gives the wearer more flexibility to adjust clothes as needed to fit a non-human avatar, or when wearing multiple mesh items. It is fairly easy to calculate. You are already calculating the deform amount for each vertex going from the standard shape to your actual shape, so strength is just a multiplier to those amounts.
-
Link: https://jira.secondlife.com/browse/STORM-1716
Assignee is Qarl Fizz (formerly Qarl Linden)
It will not have strength or bias modifiers, but it appears it will have an "on or off" boolean, so that items that should not deform can be flagged to stay the same shape. For example: A knight's shield made of mesh would want to stay the same shape despite appearance changes.
-
2
-
Facebook
Instagram
Twitter
YouTube
Flickr
small mesh object vanishes when camera is close
in Mesh
Posted
Camera near clipping has been in SL since I started 5.5 years ago. Find the oldest viewer you can, and use CTRL-ALT camera movement to put the camera just behind the avatar's head, and part of the head will disappear. Experimentally it appears to be set to 0.2 meters or so.