Hopefully you have been following the Mesh Deformer and developments around it. Today in the Content Improvement group meeting I brought up the subject of Consensus on Mesh Deformation Tools.
Some people have the same concern I do. What’s best and how do we know? There is speculation on what is the best way to deform mesh so clothes fit. Discussion extended into the work flow of making clothes too. Maxwell Graf is keen on the idea that any solution has to provide a reasonably simple work flow. I agree. Making mesh clothes is already complicated.
Karl’s Mesh Deformer is reasonably simple. It certainly avoids adding complication to the clothes making process. In general it will simplify what needs to be done to get clothes to fit. Simplicity is a large factor.
Discussion thrashed out what one would need to do to make clothes using what we call Morph Targets, which is different than Redpoly Inventor’s Collision-Bone method. Geenz Spad provided his take on what would be needed to implement the use of Morph Targets. It will likely take some changes to the asset data, the servers, and the viewer. So, it is not a free process and the Lab will be involved.
However, Morph Targets are already in use. The Appearance Sliders are activating Morph Targets in the existing avatar. In Blender morph targets are labeled Shape Keys. If you read my tutorial Working with #SL Male Shape you’ll see Shape Keys at work in Blender.
Shape Keys are created by editing the base mesh. One actually models the shape by moving vertices. One can use Sculpt Mode to build the new shape. I’m not sure whether a complete set of new vertices positions is created or not. But, it works to think of it that way.
In the tutorial I don’t focus on it, but one can use the Shape Key slider in Blender to morph between the base shape and the target shape. This is pretty much what the Appearance sliders in the viewer do. So, it seems there is no massive development project needed to create an ability to use Morph Targets. In the viewer we would probably pretty much be riding on existing code.
How we get the data into SL from Blender and then store it in the asset servers is another issue. Geenz is going to talk to some Lindens about how easy or complex it will be to implement.
The challenge here is in learning how to use Shape Keys/Morph Targets (the name depends on your 3D modeling software). Instead of… or rather in addition to weight painting we will need to model the mesh of our skirt or jacket into additional shapes. Then we have to get the weight painting to work with the shapes and the ‘tween shapes too. It is starting to sound complex.
With the added complexity comes the ability to have more control over the deformation of a jacket or skirt. I imagine I could do something about that horrid sketch section in the front and the back of skirts. So, we get more control and more complexity. Max points out this choice makes it much harder for non-professionals to start making clothes in Second Life™. That is a serious down side.
The result of the meeting discussion is a proposal is probably going to be prepared and submitted to Linden Lab™ regarding implementing morph targets.
Problem
At this point the issue for me is whether Karl Stiefvater is going to wait for this process to work its way through the Lab before doing any more work on the Mesh Deformer, which we’ve been discussing and working with for months. Karl has said he is waiting for some consensus, presumably from the community.
The next question popping up is: why can’t we have both? While a hard vote was not taken, the consensus at the group meeting was obvious that most if not all of the 30+ attendees was for the Mesh Deformer to be completed and the Morph Targets/Shape Keys proposal to move forward too.
We have no way to figure out, at this point in time, if they will both work together and be complementary or be mutually exclusive. Geenz was probably the most technically competent in this area that was at the meeting. He doesn’t see anything that makes the two mutually exclusive. So, the community consensus seems to be: move ahead on both.
Approached another way, it is also obvious that the group wanted whatever could be completed and implemented as soon as possible.
So, hopefully Karl will get moving and complete the Mesh Deformer. A couple of people asked if they could message Karl on his blog. Well… yes. So, stop by his blog and leave a note. QLab
Summary
The choice of 30 or so people is a pretty small sample, even if it high in content creators. So, take the following poll and add your voice. Contact anyone that you think should have a voice in the decision and ask them to vote.
Please, if you don’t understand the issues, don’t vote. Thanks.
Do we hold up on Karl's Deformer while we decide on other possibilities?
- No, we need it now. (51%, 36 Votes)
- Yes, we need to explore all possibilities. (21%, 15 Votes)
- Yes, we want only the best. (14%, 10 Votes)
- I don't know, I don't care, just make something work. (14%, 10 Votes)
Total Voters: 71
Do you want Karl's or Redpoly's Deformer?
- Karl's (63%, 44 Votes)
- I don't know (17%, 12 Votes)
- Red's Collison Bone (10%, 7 Votes)
- I don't care (10%, 7 Votes)
Total Voters: 70
Do you want Karl's Deformer or Morph Targets?
- I would prefer both. Karl's simple one and the complex Morph Targets later. (74%, 54 Votes)
- Yes, Karl's is enough. (21%, 15 Votes)
- No, Morph Targets are better. (5%, 4 Votes)
Total Voters: 73
Great and timely article, as usual! I would like to add a small correction:
“Max points out this choice makes it much harder for non-professionals to start making clothes in Second Life™.”
should read:
“Max points out this choice makes it much harder to make clothes in Second Life™.”
Any addition to an already extended pipeline is not a solution that should be considered. Any tool which is being considered should be prefaced by examining the amount of work which would be necessary for the designer, already bogged down with production and market considerations well outside of design, to add to his/her workflow. It is not merely a technical problem, it is also a very much a human factors issue, which is largely left out of the debate. How hard will this be to use? Does it add 2 hours to each item I make? 4? When does it stop being a solution and become a different problem to deal with?
First question:
Do you understand how morph targets work, and have you ever worked with them in your 3D modeling pipeline.
If you answered “no” to the above, than the second question is for you:
What makes you think that this will make this much harder to make clothes for Second Life?
I honestly don’t think you understand how morph targets work Maxwell, and rushing in to pass judgement on them like you seem to have done thus far seems a bit bone headed. I’d suggest doing a bit more research if I were you; on top of that, they most certainly wouldn’t be any worse than the cBones idea.
I think Max understands. One measure of complexity is to look at the number of video tutorials uploaded to the net in the last year. Compare the number of other modeling video tutorials added. I think the number of tutorials is influenced by the simplicity of the subject.
While I see morph targets/shape keys as giving me more control, I see it also requiring more time and effort. Which is why I want both the Deformer and the Morph Targets. I see the Deformer as being way easier for new modelers to design for.
The Collision Bone process Redpoly revealed would also add more control but be more work.
I don’t see how you make a conclusion from there. I think there’s roughly the same number of tutorials on putting down a simple box and on rigging a humanoid character, maybe with a few times of difference. Also i don’t understand how you mean it, in which direction. Let’s say, rigging a character needs 100 times more skill than rezzing a box – i think that’s reasonable, because few people manage to rig their first humanoid character in a satisfactory way before spending a dozen hours with the program, while they will probably have a box rezzed mere few minutes after starting the software for the first time. Now, given that assumption, would you say there should be 100 times more tutorials on rezzing a box, or 100 times more on character rigging?
So let’s see how it works out:
“blender rigging tutorial” on youtube: 3,670 results
“blender crate tutorial” on youtube: 100 results
“blender box tutorial” on youtube: 1,410 results
“blender cube tutorial” on youtube: 5,410 results
Note however, two last queries are very broad, and produce results mostly hardly related to cubes or boxes, just see for yourself. At least a whole lot less specifically related than results about rigging.
I believe that number of tutorials is not majorly influenced by simplicity, it’s influenced primarily by popularity of the subject. Something utterly trivial yet something noone desires is likely to have no tutorials at all. It’s also related to what is the minimum necessary knowledge or work needed to obtain a goal. For example, if you want to animate a character in Blender, you need to know rigging, but you don’t necessarily need to know shape keys (morph targets). Now, for characters with subtle or complex facial features, you will certainly receive better results if you use shape keys, but since it’s not absolutely necessary for the start, many recent tutorial series don’t contain tutorials on that – but this used to be different back when skinning was a lot less evolved and animation system wasn’t layered, so shape keys were often taught before rigging.
Good logic and reasoning very well expressed.
My take on the reason tutorials are made is that a large portion of them come from those that had a challenge and assume others will too and thus make a tutorial they feel is needed and helpful. But… it has to be something they have learned and feel confident enough to make a tutorial. I see a lack of information on morphing. But, I could be mis-taking what I see. It may well be more of a personal perception than hard numbers. Still it seems apparent to me.
Of course the search terms can have a major influence on the search results so that has to be handled.
I do think forums like CGCookie makes tutorials based on popularity and perceived need for the tutorial. I factor those issues in my head and have a more intuitive sense than the typical male linear sense. The advantage to linear reasoning is it is easy to document and debate.
The real objective measure of whether morphing is adding significant work at greater complexity comes from doing it.
I am building things to support future work and experimenting with various ways of making mesh clothes. I am acutely aware of what I have to learn to get some of these steps to work. From that point morphing does seem to add some significant complexity and effort. I can imagine a time when I have the tools and learning in place and will then think it is relatively easy, which I think many that know 3D modeling forget and overlook.
Actually that’s not nearly as much of an issue as you think. Yes, manual creation of shape keys/ morph targets is an extra step. However, with Qarl’s mesh deformer, the clothes creator will very well be need to be aware of the behaviour of deformer in problem areas such as near creases, requiring extra work to learn and avoid, as well as longer iteration times on finding clothes designs and topologies which are less prone to such issues. So it’s not clear which is more work, if the goal is to create high quality content. As for creating morphs, yes, it definitely is a major piece of extra work, but with good tools it’s rather simple.
And finally, morphs don’t need to be generated manually. If morph target solution is implemented, it is absolutely trivial to adapt Qarl’s deformer to generate a full set of morphs, which would really bring the consumer and the designer effectively the ultimate choice. Also it would allow to develop more elaborate fitting algorithms instead of being stuck with one forever. Qarl’s or other such algorithms can then be integrated into the viewer and used at upload time. Viewer developers and independent tool vendors can rival on providing the best possible automated morph generator, or provide niche morph generators with special features for specific kinds of content.
Finally, one possible implementation strategy involves only partial modification of the viewer – for worn meshes, a different mesh asset per avatar with morph state baked in could be generated on the server, requiring the wearer to possess an up to date viewer, but allowing legacy viewers (with mesh support) to render the content, allowing for rapid deployment and smooth transition.
Major grounds on which one could object to morph targets, are technical. It needs to be worked out how to keep bandwidth consumption and memory consumption client-side to a minimum, especially in crowded settings. Also, the increased amount of initial data could prompt LL to make upload of morph set very expensive. Other problem is that the process is behind closed doors between Geenz and LL.
Saying the process is behind closed doors may be a bit misleading. The Lab is not going to publicly talk about some of these issues and ideas. If Geenz can get information in and out and the Lab is comfortable trusting his discretion to relate their comments and ideas only as appropriate I think that opens the doors a crack. And.. it is a crack they aren’t generally open.
I believe I’ve openly invited you to help come up with a solution to the bandwidth problem, Siana.
Both Daz and Poser are extremely successful packages oriented around fitting mesh clothing to figures. Morph targets are a big component of how they get things to work, though not the whole of it.
Their communities are large and very vibrant and also have a lot of people using Blender etc. to generate/modify content. They get some very spectacular results. I imagine there has to be at least some overlap with the Second Life community. Have people from their camp, knowledgeable about the ins and outs, weighed in on this discussion at all?
Perfection is an ideal, not a goal that can be reached. If we stop Qarl to send him on the track of this illusive perfection instead of allowing him to finish the Mesh Deformer (as imperfect as it is), when will we see anything arrive in SL at all? How long has been Qarl working on the Mesh Deformer already?…
Yes, it’s not ideal but it’s a step in the right direction. It can always be improved later.
A little something is better than nothing at all. Right now, meshes are good for nothing but static props. Clothes don’t fit. Scripting is very limited. Let’s just improve this situation. Let’s get Qarl’s Mesh Deformer. Now! Not next year or in some unforeseeable future…
Riisu, yours is an understandable and respectable opinion. I will point out a flaw in your thinking. No one is stopping Karl or trying to divert him. The apparent majority want his work completed. Also, he has been paid to make the Deformer, not morph targets nor collision bones.
Pingback: Mesh clothing deformation: options | Living in the Modem World
For question #3: I want a pony too.
Of course, I *want* to have all the options, to make things as good or as simple as I care to.
But I think Qarl’s deformer is *good enough* that the resources which would be required for morph targets are better spent elsewhere.
We disagree. Except for user community resources there is unlikely to be any diversion of Lab resources near term.
The Lab is going to set priorities for their resources. We can voice our preferences and I think should to provide them some indication of what the community wants. The Third Party Developers will do whatever they want. So, we can’t really divert them.
I have a hard time believing that developing a morph system from scratch will not tie up more resources than plugging in the almost-finished deformer.
It may be resources well spent (and the combination of autogenerating morphs with the deformer and taking over manually when needed which Siana mentions sounds pretty sweet), but my point is mostly that a poll like this does *not* really give any indication of actual preferences. Who’s not going to click “Of course I want all the cool features”?
The morph system is already in use in Second Life. So, ‘from scratch’ not completely accurate.
The people that are NOT going to click on ‘I want both’ are typically those that are thinkers. Of course in this case thinkers may be clicking ‘both’.
The idea I express is that we can have the Deformer and move on to Morphs. Today’s article on Content Improvement expands on having both ideas.
Semi-commenting on that later post, yes, I like the possibility of both, and the focus on the manageable learning curve.
Personally, I am generally of the “give me all the power tools, and I’ll figure out how to use them” mentality, but SL thrives on the “jump in and play, and see something work” attitude, and I think it is important to keep that.
You phased that learning thing way better than I did. Thanks.
Naturally, morph targets would be a ways off. Qarl’s deformer would most definitely be a much more immediate solution; and there’s no point in shifting resources around to accommodate the development of something that probably won’t start serious development until the end of the year at the very least.
Qarl’s deformer enables automatic results, and it’s quite close to being finished. For many content creators that just don’t feel like learning how to use morph targets, this would be the best option for them. Morph targets will always be a “later down the road” goal from my perspective, until the mesh deformer is implemented at the very least.
Pingback: Making a Difference | MetaReality Podcast
I believe that LL’s track record proves that if they put out a feature, like the Karl Mesh Deformer, they will shelve the other projects like the RedPoly Mesh Deformer or Geenz Morph Targets, because they will have solved the ‘problem’ and the rest will seem like merely ‘icing on the cake’ or , in other words, ‘non-essential’. They will see it as time to move on to some other ‘problem’ with their time and resources. I voted to have all of it because I don’t want to see them drop the other stuff just to get out the easiest thing.
I disagree. One change in the Lab is the proposal process, which seems to work now.
It looks like the the Deformer is going to move forward. It now seems to be a matter of keeping Karl and the Lab talking.
Geenz is helping the Morph Targets idea move forward. I expect it to advance to a formal proposal before the end of summer. Creating either Collision Bones or Morph Targets will be mostly work done by the community. The changes on the Lab’s side appear to be rather small and simple. So, the idea we will be limited to ONLY one or the other seems unlikely.
Pingback: Dies und Das aus dem Web (KW 28) - [O.L.D.] Old London Docks
Pingback: Alternative to Mesh Deformer? - Page 4 - SLUniverse Forums
Pingback: Content Improvement Week 28