Loki's Adventures in creating a mesh object for SL!

[Loki Eliot]

Thanx for the tips, it turned out i had Mirrored my object at some point inside out. Mirroring back seemed to have fixed that but it meant i had to re-rig left and right body parts again.

After uploading the Costume was the right way around but .. well as you can see , spikey. I read or saw somewhere on machinimatrix about spikey rigging so i shall look into that again

Im almost there !

 

[Drongle McMahon]

Mirroring in Blender does flip all the normals*. You can flip them back again with Mesh/Normals/Flip. No  need to invert the mirroring. Too late, as you did it already. :smileysad:

* Strictly speaking, I suppose that is not totally accurate. In fact it simply fails to mirror the normals while it does mirror the faces. This has the effect of turning inside out.

[Luc Starsider]

I came across this tutorial on weight painting in Blender 2.5, and I thought you might find it useful (perhaps others will too). Although it is not quite the same as my own workflow, it gives a couple of *very* useful tips, I think, that is going to help with the process.

Regarding the spikeyness in you mesh. It is either, I believe, because your modifiers - subsurf and perhaps also the mirror - need to be applied before rigging/skinning/weight painting, or it is because there are som stray vertecies that are weighted to the wrong bone, not weighted at all or belong to the wrong vertex group.

- Luc -

[Loki Eliot]

Ok the final problem was that i had not assigned all the verticies to a limb on the simplebot rigged skeleton. This either made me spikey or not upload at all.

Right, so i learnt alot while trying to achieve rigging mesh. The hardest part did in fact turn out to be rigging. Most of the pages ive read make it sound like its simply a third step but the reality is that you have to learn how to assign vertices to parts of the body, and alot of care is needed to be learnt on how to efficiently and correctly rig parts of your mesh to the right parts of the body.

Although not very detailed my wolf costume has shown me just how huge the workflow and process is to getting Mesh clothes created. Now that im aware of it, i will now move onto making a pair of shorts and a shirt to see if there is a massive improvement that can be achieved over standard default SL clothing.

Thanx everyone who helped me on the forums and i hope this Topic helps others.

[Vivienne Daguerre]

Looks good! I must say, you seem to be picking it up quickly.

[Drongle McMahon]

Thank you for the story. It will be a good encouragement to others.

[Loki Eliot]

Not quite over yet. next i want to see if i can make actual replacements for T-shirts and pair of shorts using the skills learnt over last week or two. Its more of an experiment to understand what sort of products could be made and whats needed to give others mesh clothing.

For instance this first t-shirt i created. I've learnt that you would have to be an absolute expert to mesh a shirt over your Avatars mesh without bits poking through. So when selling a mesh shirt you also need to sell  Alpha Clothing to wear under it. Alpha clothing  as Torley

makes part of your Avatar mesh invisible. When making Mesh clothing that will go on top of your avatar mesh, people will also need to learn how best to Alpha their avatar mesh underneath... i think that all makes sense.

So Goals are.

1. Make a Mesh T-shirt.

2. Make mesh pair of shorts

3. create Alpha clothing to wear underneath.

4. give to someone else to wear.

So this is the T-shirt, created in Blender 2.5, Rigged in Blender 2.49 and Texture created in Photoshop extended edition. I used a small tutorial on how to bake an Occlusion map (renders shadows cast by your object for use in texturing).

As yet i have not made my pair of short in mesh yet

As you can see i have not yet created Alpha clothing to wear under the T-Shirt. Thats my next  area of experimentation. One thing i have noticed is how the Mesh Shirt maps the rigged parts to your avatars body. Where you have rigged say the pelvis to the torso depending on your avatars shape settings, things can look a bit bunched up. I need to learn how to combat this if i can at all.

[Luc Starsider]

It's impressive what you have managed in such a short while.

I think the edge/face flow in the mesh, and the density of the mesh in the defomation areas is key for rigged meshes, as well as getting the weight painting absolutely right. Even then, chances are you're going to have to use alpha layers to hide some bodyparts. Reading, learning and practicing are the things to do to get better at it. This is what I'm doing myself these days.

A new avatar mesh with better topology would probably help in this way, but it does not look like a new mesh is on the table, so we better learn to work with what is available atm.

- Luc -

[Loki Eliot]

"edge/face flow in the mesh, and the density of the mesh in the defomation areas is key for rigged meshes, as well as getting the weight painting absolutely right"

Hugh????:-p yeh you might need to explain what all that means hehe

[Luc Starsider]

Sounds impressive, doesn't it.

I'll try to explain. You have most likely thought of this yourself in the process, I'm sure, and others can most likely explain things way better than I can. I feel I should say that I'm not that good at these things. I have read a lot, watched a bunch of tutorials, tried many things (and mostly failed in achieveing what I want). Thus, I have many of the technical aspects relatively clear in my head, but not necessarily the skill to use them in the best way. I'm saying this so you know where I'm coming from, and I hope I'm not to off the mark with what I say here.

Edge flow means simply how the connected edges flows in and along the model/mesh. You see this when you alt+right click an edge, and the loop is selected from start to end. Perhaps it is a complete loop around (part of) the model, but perhaps it is not. I assume you know all this; I just felt I had to start with it still.

When modeling for animation, the result thend to be better whan these edges follow what is natural for the model. That is, if you model a human - or clothes ment for a human in your case - you tend to achieve the best results for animations if the edges follow (to a degree) along with the muscles in the body. Which in turn have an impact on how the joints bend. Because of this, it is sometimes (often?) necessary to take control of the edges and change them so that they flow the best way for the task at hand.

By 'The density of the mesh in (perhaps around would have been a better word) the defomation areas' I mean both how many loops you add around the shoulder joint, and how far apart they are. For instance; one loop defining the shoulder is not letting you control how  the shoulder deforms when you rig the model and move the arm up and down. There simply is not enough vertecies to assign weights to to make it move convincingly. Two loops give you more control, and you can sort of paint weights but it is still not going to look good. From what I've seen, what seems to be the established standard (for lack of a better term) is to use three loop defining joints which give you enough control when weight painting to make the mesh deform convincingly. (Also from what I've read and heard, the shoulder is one of the most difficult joints to get right.)

Whether any of this makes sence, or is correct, when you try to attach a rigged mesh to the SL avatar is another matter. Perhaps other techniques and considerations are best for SL, I don't know.

Anyway. I hope this make sence, and I very much hope I'm more or less right in my descriptions regarding this. If I am wrong about some or all (*shudders*) of it , I hope someone comes along and provide the correct information. And, also, forgive me if this is all known infomation to you. Hopefully you get something out of it.

[EDIT] The tutorial on weight painting I linked to in a previous reply probably shows all of this much better than I'm able to explain it.

- Luc -

[Loki Eliot]

OK so i took a look at that video on weight painting and i think i understand more about make the joints in my rigged mesh more organic. So thats a shirt done, next up is the pair of shorts.

A friend asked me what happens when someone with a bigger avatar wears the same shirt. The Answer is the Mesh scales to avatar size. There are a few discrepancies witch need fixing but otherwise one size fits all

I also noticed today that the latest Mesh development viewer for Mac no longer does shadows and depth even though all the tabs are clicked.

[Loki Eliot]

Ok its been a while. During the last month i got as new iMac and LL fixed shadow, lighting and general performance with the mac viewer, Kudos to them.

I still need to perfect my t-shirts and start on a pair of shorts, but in the mean time ive been playing around with how MESH can be used to create environments.

In many computer games the surrounding buildings are low detail and you cant enter them, they are for show an create atmosphere. Using Blender i created some real basic buildings and textured them in photoshop.

I made about 3 different shaped Buildings each with one 1024 texture. I also created a cobbled street and started placing the buildings around it to create a tudor style village. It's amazing how suddenly atmosphere appears after placing a few objects, the magic of creation .

I also looked at creating a cave with ONE mesh. One of the things im most interested in is how Mesh can be used to create more natural enclosed spaces such as caverns. The result was a floating rock which was hollow and filled with crystals.

A single hole allows you into the cave. Here i ran into something that is either a bug or i lack understanding of. Its inplied  during the upload that Convex Hulls are more accurate and efficient form of Mesh Physics and there is an option to have the mesh holes CLOSED when uploading. At the moment it seems during upload all holes in a meshes physics are closed no matter what. The only way to allow your Mesh to have a hole in it is in 'features' tab set from 'Convex Hull' to 'Prim. This allows the holes to exist but the trade off is that the Physical shape of the Mesh is less accurate.

Anyway here is a video of my Mesh Tudor Village and the Floating Rock. Total Prim equivalence count for the whole setup is just under 200 prims. But could be less with more practice.

You can visit this project here secondlife://Aditi/secondlife/Mesh%20Sandbox%207/88/92/23 on the beta grid

 

 

[Vivienne Daguerre]

You are talented! I look forward to see what you create next.

[jimbolimbo Oval]

This is Amazing Loki, I never thought i would see builds so detailed and intricate as what you have created, You have really opened my eyes to what meshes can do! I know ill never be able to branch into then (Heck, all i can do is create a simple prim box.) But It's amazing to see you're builds and I'm so glad the Kids community has someone as talented as you building, As i can see that when this becomes live Escapeds it going to make some big changes!

[Loki Eliot]

Today i started working on converting my old Airship into Mesh to see how it would work. Already I'm struggling to get a decent result with the Balloon. Im not sure if i've just hit the limitations of SL mesh or if im just not modelling very well with Blender.

Basically it seems IMPOSSIBLE to create a Mesh equivalent of a Sculpty Balloon. I used one Sculpty to make the Ballon for my Airship a year ago. Trying to replicate it in mesh is proving impossible, it's less detailed and more blocky yet costs more Prims. So either im lacking vital efficiency in modelling or in this instance the Sculpty wins.

It's rather confusing in a way.

Either it's all totally gone Whacko, or Linden Lab have done this on purpose to significantly drop the detail in Second Life to improve performance. If we all recreated everything in SL with this low detail im sure a mobile viewer could render it no trouble LOL.

[Gaia Clary]

You can get the costs down by linking multiple mesh objects together.
Also you can get the costs down by adding the simplest possible physics shape.

Currently big objects get realy hard punished by mesh PE. It is not clear why this is so. And the question why a BIG low poly high efficient mesh costs more than a BIG high poly low efficient sculpty, is only partially answered so far.

Lindens ask us to not make this comparison. Although everybody does and (imho) it makes a lot of sense to compare ;-)

[Loki Eliot]

Im just comparing a mesh that looks crap and costs more than a sculpty that costs less and looks better, how can i NOT compare that???? LOL

[Loki Eliot]

I'm not a 3D modelling expert yet so im not at all understanding whats going on with Mesh at the moment. 

I decided to do a test as im trying to replicate a Sculpty/Prim based Airship with one made from Mesh. Really i would like to make it with less Prim Equivalence or even settle for the same amount, but it's not looking at all like thats possible so far. I shall persist.

Here i have taken a sculpty of a Ships wheel, 1 prim. I saved the sculpty texture to my desktop then imported to blender, exported to Collada then straight back up to SL as a mesh. During the journey the Ships wheel gained 4 PE.

See in my simpleton experience, surely they are the same geometry, same amount of faces and should cost the same? The Adventure continues....

[Gaia Clary]

1 prim meshes can not be made cheaper than 2 PE. Meshes get less in price when bundled into linksets. Dont ask me why this is so, but it works. Also the problem which makes meshes much more expensive typically comes from the default physics shape which is way(!) too detailed.

If you can create a simple physics shape with just a few triangles, you get physics weight down below 1.
if you create very low triangle lowest LOD your download costs go down significantly.
if you can bundle multiple meshes into one linkset, your server costs go down significantly.

Here is my current project:

 

Sculpty object (PE=4, 8000 tris for LOD3):

Comparable mesh object not fully optimized (PE=4, 4000 tris for the sum of all LODS of all parts)

Closeup of mesh version:

In this sense you can make comparable objects in price (as long as they have less than a few meters BBox side length). For larger objects the download costs raise beyond usability (as you have already observed)

[Gaia Clary]

So instead of talking, i prefer to show ;-)

 

• Check it out at: secondlife://Aditi/secondlife/Mesh Sandbox 20/250/196/23
• The wheel has a diameter of 2.46 meters, id say a bit big for a ship wheel ... ;-)
• The PE=2 comes from the Server Weight, which can not be pushed below 2. So whenever you rezz one object the minimum costs are PE=2. But if you add the wheel to a link set, its costs go down, e.g. 3 wheels in one link set -> PE=3
• Also note, that this wheel has 4 separate texture faces (marked in red, brown, yellow and green).
• Furthermore look at the wheel closely. You see that it has some smooth part and some flat surface part. This is because we can tel for each face if we wanted it flat or smooth. With sculpties you would need to collapse (waste) rows or columns to get the same effect.

Now compare THAT with a single sculpty please ;-)

 

[Loki Eliot]

You show me the result, but not how to get there. There is so much technicality that i still dont understand.

[Gaia Clary]

Basically i created the wheel from scratch. I did not derive it from a sculpty because that would give realy ugly results:

 

1. The outer wheel (yellow): A torus with 32  * 6 faces around
2. The inner wheel (green) : A circle with 8 vertices, filled and extruded along y
3. the spikes cylinders with 8 faces along circumference and one face along the long axis. Both ends extruded to the red handles. There are actually 4 spikes in this mesh, the intersections are hidden in the green inner wheel. That is realy way cheaper than creating 8 of them ;-) The handles and the spikes are on one single mesh, but use 2 different materials.

For the LODS:

I removed edge loops where they would not harm for the next lower level of detail.

For LOD0 special case:

I cheated a bit, but it works in world: I reduced the LOD0 to 2 circular surfaces. I could then use a billboard image. Actually i made a mistake in the object and i used 15 faces, where i could have done the same with only 8 faces. This would drop the download costs even further down. And since i also used the LOD0 as physics mesh, i could save another 50% of the current physics costs (1.0)  and so i probably could make the wheel even bigger without getting above PE=2

Note: I did not do any unwraps, so textures can not be applied to the object (only color). Doing the unwrapping would give you the chance to make this look realy good and detailed. But it would take another 2 hours (for me) to get there.

You need it more detailed ? You could figure out by yourself, or wait for the tutorials to get ready ;-)

 

 

[DanielRavenNest Noe]

They are the same geometry at the highest level of detail.  How they are reduced with distance is different.  With sculpts, my understanding it is by deleting every other row and column to get a map half the size.  With mesh it is whatever you design if you make custom LODs.  If you use the auto-LOD in the uploader it is whatever it generates.  In general a poor physics shape will impact the cost a lot.  It is costly for the simulator to calculate collisions between objects with detailed shapes, so that is penalized heavily.

What you discovered with the balloon is the prim cost for mesh is weighted for size at approx the square of the object's diagonal.  So beyond a certain size, they get very expensive.  Large objects get seen at higher detail at longer distances, so the cost of mesh reflects that.  The cost of prims and sculpts *do not*, so they tend to win the larger you make stuff.

[Loki Eliot]

I appreciate all your advice and encouragement. Is their a tutorial that explains LOD and how to prepare for it when uploading a Mesh to SL? it seems a real vital part of Mesh, yet i don't really understand it and im sure others just starting out never even heard of it.

[Gaia Clary]

i am in the middle of making part III of my tutorial series. If you like, you can checkout the first 2 parts of the

kettle Quest tutorial

Part III is exactly adressing how to create the LOD's for meshes. it will take me another 10-20 days until that part will be good enough for release ;-)