3D-Coat 3.3 updates thread

[Psmith]

Just for the record, while working the booth at Siggraph, I spoke with a veteran game designer/project manager with a long career at Epic Studios. He is now launching his own independent game studio which, I believe, will rise to the top of the gaming heap. One of the first things he said, after watching all of our auto-retopo videos, (on the big screen that Don brought) was this:

"It is just amazing how much time modelers waste being obsessive over perfect topology. It's one of my biggest problems running a studio. You can't stop these guys, (who are highly paid) from spending countless hours on topology issues that just don't matter. If you can't see the result of fudging in the game itself, then adding perfect topology doesn't benefit anyone. Right now, I'm looking at 3D-Coat as a means to outsource all of my modeling tasks to China, where I can use relatively unskilled modelers, coupled with using 3D-Coat, to lower my costs substantially. This new auto-topology will enable me to do this."

If the person I quoted is reading this, please forgive the paraphrasing - I know it's not exact, but the spirit of the thought is there. I'd love to do an actual, in depth interview with you where we can tackle, and lay to rest, this annoying and ever present argument once and for all.

So, from a top industry veteran and professional - from the horses mouth, itself, realize that this is what matters regarding topology: Give me any tool that will produce satisfactory results at the lowest possible price. This is how the new and successful gaming companies think. This is how the managers and proprietors think. This is what matters.

And don't throw out the argument that WETA modelers would not be happy with these new features. WETA makes their own tools to do very specific jobs. So do the other large studios. In fact, you would be completely surprised to discover how many of these large studios use purely algorithmic solutions to solve many of their modeling and animation problems - not solutions manually executed by "professional" modelers.

Stay focused and obsessed with perfect modeling topology if you want to, but the industry, itself, will not find you a particularly useful asset for their most urgent needs.

Greg Smith

[splodge]

Stay focused and obsessed with perfect modeling topology if you want to, but the industry, itself, will not find you a particularly useful asset for their most urgent needs.

Greg Smith

Depends on who you talk to in the industry. And just one person isn't the industry.

People have different requirements. I'm pretty sure that there are art directors that would reject a model with poor topology.

For artists to get employed at a games studio it's essential that they can make nice models. If they had submitted models with poor topology then they would never have got hired.

For many 3D artists making a model with a good topology isn't that difficult as they box model right from the start.

There's a lot of artists that like to sell their artwork on Turbosquid etc. It's important that the topology looks clean because many buyers are very aware of the problems that can occur with a poor topology. So these buyers will only buy models that are listed with screenshots showing wire frames.

Now if you're somebody that's thinking of going alone and making your own game then it's entirely down to you to decide on how good the topology needs to be. Although don't be surprised if game reviewers complain that your game is graphically crap and has poor frame rates.

A nice topology means nice shading and lower poly counts.

[Psmith]

A nice topology means nice shading and lower poly counts.

Yes, you are right about this. But, Andrew's auto-topology is not poor topology, but actually very nice, and it is getting better by the day. Low poly counts included. What is salient to the industry is cost effectiveness - not perfection.

And, as budgets get tighter, cost becomes more and more of a consideration. Let me also add that, at the same time, fewer and fewer "professional" western based, (highly paid) modelers will find themselves employed by anyone.

Making the suggestion that something auto-topologized in 3D-Coat represents "crap" topology is simply ridiculous and a misrepresentation of the facts. And, right now, all of the facts have yet to materialize themselves out of Andrew's most fertile imagination.

Greg Smith

[splodge]

Making the suggestion that something auto-topologized in 3D-Coat represents "crap" topology is simply ridiculous and a misrepresentation of the facts. And, right now, all of the facts have yet to materialize themselves out of Andrew's most fertile imagination.

Greg Smith

Hey! I wasn't making that suggestion. The topology from the new auto retopo is looking like it's going to be very nice.

[ghib]

*This isn't intended as a post of dissension*

I'm all for Auto-retopology but here is my limited experience with it so far.

I attempted to use auto retop on an asset to be rigged recently (just as a test) and wasted a good deal of time fixing up the result where it didn't get it quite right. This was even with the placement of suggestive strokes.

I could have retop'd it much faster doing it manually; as I have been using this method of modelling for a while now and have an eye for where to place the loops. Much like a lot of artists now I imagine.

The reason it took so long to fix up the 'bad' topology is that the tools are all still separated and you cannot paint select/delete polys & edges.

I would highly recommend anyone having a look at 'Quad Draw' in NEX for Maya; which is like a more sophisticated version of 3DC's 'Points & Faces'.

It has elegantly placed most modelling functions in the 1 tool and uses a combination of sticky key presses to activate them. i.e. press and hold CTRL to paint delete polys/edges/verts.

There are too many subtle workflow adjustments to post in this thread but they have been well documented in the suggestions thread already.

[kay_Eva]

I appreciate your point of view ghib but 3dcoats autotopo is still not complete so it's a bit misleading to compare it's current implementation to anything else

and besides that, other than the symmetry issues and the fact you can't use it on poly objects yet (only voxel objects)... I'm pretty much in love. I retopo'd a human figure in about 30 seconds. It wasn't perfect. But it was good enough. Seriously. In the ways that it counts, the joints and everything were deforming just fine. There were no problems. but I'm hardly a pro so take that with a grain of noobiness

[stevecullum]

Give me any tool that will produce satisfactory results at the lowest possible price. This is how the new and successful gaming companies think. This is how the managers and proprietors think. This is what matters.

I think what matters is understanding where and when you can take short cuts with topology. When it comes to animation, good topology most definitley matters, so while auto solutions can provide a starting point, its no substitute for someone who knows how to model cleanly with correct poly flows. I suspect the guy you spoke to, will almost certainly need to employ someone with good skills to 'manage' the flow of models coming from China and make tweaks and changes to them for use in the game engine. It merely provides a means of reducing the number of skilled modelers required in his business, to make it more competetive.

[Psmith]

Actually, we spoke at length and he was very clear in his response to what he saw being demonstrated regarding 3D-Coat's auto topology - (it's current permutation in conjunction with what Andrew plans to add in short order). The next features would allow a modeler to paint weights indicating quad density, provide accurate edge snapping - generating a finished mesh of nicely arranged quads.

He felt that with those added features, most of his game models could be created in the voxel room, from scratch and auto-topo'd, (not a real word), with very little need for any highly trained modeler to do anything "post production", so to speak. He also commented that all the worry over using things like triangles and diamonds in hip areas is overblown and that these things don't matter if they don't show in a typical game character's mid-distance animations, (which constitute most of the animations used). I believe that triangles won't even be an issue after Andrew has finished this current phase of development.

So, the gist of our conversation was that he could very quickly train relatively inexperienced modelers, (inexperienced with perfect sub-D and retopo work) to complete all the modeling and texturing tasks, from start to finish, all the way over there in China without the need to employ any high paid "western" technicians for that part of the work. At least he thought so, judging by what he saw and heard at the show. Certainly there would be someone, somewhere in the chain that would oversee all completed work by everyone.

Really, the positive response from all of the real professionals that I spoke with was overwhelming. We were absolutely swamped, for most of the show, with more interest and enthusiasm than we could adequately respond to in the time available. Everyone employed in the business that I spoke to also indicated that it was such a relief to see these automatic features finally become "real", since they all hate the topology and UV parts of character, prop and set development.

Greg Smith

[AbnRanger]

I saw theirs going to be the ability to paint density via freeze in an upcoming build (via twitter). If it hasn't already been suggested, would be nice to have a density-by-angle auto-paint feature. So if the surface is more curved it'll be denser then a flat area.

That is actually a VERY good idea. Glad you brought that up.

Andrew would it be possible to have 2 separate modes? One is Paramaterized, as it is now (evenly-spaced polys) and one an Optimized mode, where a separate modified algorithm is used....a surface with flat or very low incidence angles get a greatly reduced number of edgeloops, and high incidence angles get more? Also, any surface with an 80-100 degree change in angle automatically gets 3 edgeloops at that apex (to preserve the hard edge, of course).

An OPTIMIZED mode would obviously be ideal for animation and low-poly game assets, with the other for going into ZBrush or Mudbox for further work, as they like evenly spaced poly's.

[kay_Eva]

Will it be possible to just autotopo a poly mesh while avoiding the convert to voxel step altogether? I ask this because sometimes converting to voxels will distort the shape quite a bit at times and I still rely on sculptris for sculpting details into a mesh. I have difficulty using voxels for details though I know others don't. It's important to me whether you can skip the voxel step for autotopo.

[michalis]

We can manually re-topo a head for example, using 100-200 poly mesh. Tri are acceptable here. Now we can subdivide, do some move mostly modifications and finish having an excellent topology in 10 min.

The problem begins when the subdivided vertices are closer to the surface we don't want to. Typicaly behind ears or in opening areas (like mouth). Is it possible the new vertices can "read" the data of the first low poly vertices? Which is the surface they belong? Its just a thought, I'm not a programer, Andrew know much better. Please don't start the pc vs mac story again, I'm not complaining here.

kay_Eva do some manual simple retopo, I know the fear, its more simple than we may think. I know that you already did re topo but a simple one as I described?

[Javis]

Will it be possible to just autotopo a poly mesh while avoiding the convert to voxel step altogether? I ask this because sometimes converting to voxels will distort the shape quite a bit at times and I still rely on sculptris for sculpting details into a mesh. I have difficulty using voxels for details though I know others don't. It's important to me whether you can skip the voxel step for autotopo.

Yes. In the next build you can do this (assuming Andrew includes it in the next build, which I think he will be doing).

You can see from his twitter quote in the thread below, that he has added the ability to auto-retopo an existing mesh without the need to convert to voxels first.

http://www.3d-coat.com/forum/index.php?showtopic=4372&view=findpost&p=47017

[stevecullum]

Unless I've missed something, the quotes say you can import from the file menu, then it will automatically convert to voxels. So it still can't operate on directly mesh items yet. But hopefully that will change soon.

It will automatically close holes, convert to voxels...

[philnolan3d]

Yes it will still convert to voxels first. Hopefully this will change.

[MaDDoX]

Actually, no idea where on such models may be a control lines for the super algorithm, and it makes me sad. I'm afraid I'll be outside playing with that nice toy

Where to put lines in such non-organic models? Probably nowhere. Those are static models, you'll hardly have the need to deform them, so "1-click" auto-retopo should work just fine. A bit hit-and-miss in a couple corners (specially when smooth and rigid corners are mixed in the same model) as some screenshots have show, but in practice they'll be a) very easy to manually clean-up afterwards or can be simply ignored anyways since you'll probably use a normal map in your end application to reproduce the hard and soft angles anyways, as long as the major silhouette features are preserved.

Guides are only all-mighty important when we're talking about models for organic/character animation. Then things get tricky because you'll generally need the best possible edge-bracketing by the placement of poles. Poles are vertexes connected to three edges (also called Y- or N-poles), five edges (aka star-poles or E-poles) or more-than-five edges, also called extraordinary poles, although it's worth mentioning that this last type is generally to be avoided 'coz it tends to ruin the UV smoothing process. Proper Placement and connection of poles, or "bracketing", is absolutely critical for good deformations in low poly models. In the modern auto-remeshing algorithms proposed (most noticeably Bommes and Huang), the Y-poles are shown as blue dots and the star-poles as red dots. In my opinion the more control we have on the placement of those poles, the best. From the screenshots - I haven't still tested it first-hand - the crossing of guidelines are prime candidates for the generation of poles, but the lack of snapping and any visual feedback on if the generated poles will be the Y or star types hampers the user ability to fine-tune the edge flow. One possible way to improve the situation is, after a first "preview retopo", shown in the retopo room, to highlight those poles in blue and red and enable the user to re-snap them to other edge loop crossings as needed, before generating the final mesh output. This could be hard to implement but would enable outstanding control of edge flow where there's no drastic change in mesh topology and the automatic routine falls back to guesswork to define the placement of poles.

For better/more visual reference of what I'm talking about - which might be alien matters for whoever just make "casual" topology work - please check the following links and it should all be clear:

http://www.subdivisionmodeling.com/forums/showthread.php?t=907

http://www.subdivisionmodeling.com/forums/showthread.php?t=8911

To sum things up, re-topology is only really needed for low-poly related work, which include base models that will be animated (to make the process speedy and practical, since high-poly animation is still a technical nightmare) or models targeted for real-time applications like games and simulations. For static models, normal maps will make up for 99% of any weird edge flow in the recreated topology. For animated organic models, edge flow is much more important but the density is not all that important - it can be added/removed in record times with very basic modeling operations - as long as the Y and star poles are placed appropriately.

[stevecullum]

Andrew has posted file links to the latest update on his twitter page if anyone is keen to check out some new changes...

[synapsemassage]

Actually, we spoke at length and he was very clear in his response to what he saw being demonstrated regarding 3D-Coat's auto topology - (it's current permutation in conjunction with what Andrew plans to add in short order). The next features would allow a modeler to paint weights indicating quad density, provide accurate edge snapping - generating a finished mesh of nicely arranged quads.

He felt that with those added features, most of his game models could be created in the voxel room, from scratch and auto-topo'd, (not a real word), with very little need for any highly trained modeler to do anything "post production", so to speak. He also commented that all the worry over using things like triangles and diamonds in hip areas is overblown and that these things don't matter if they don't show in a typical game character's mid-distance animations, (which constitute most of the animations used). I believe that triangles won't even be an issue after Andrew has finished this current phase of development.

So, the gist of our conversation was that he could very quickly train relatively inexperienced modelers, (inexperienced with perfect sub-D and retopo work) to complete all the modeling and texturing tasks, from start to finish, all the way over there in China without the need to employ any high paid "western" technicians for that part of the work. At least he thought so, judging by what he saw and heard at the show. Certainly there would be someone, somewhere in the chain that would oversee all completed work by everyone.

Really, the positive response from all of the real professionals that I spoke with was overwhelming. We were absolutely swamped, for most of the show, with more interest and enthusiasm than we could adequately respond to in the time available. Everyone employed in the business that I spoke to also indicated that it was such a relief to see these automatic features finally become "real", since they all hate the topology and UV parts of character, prop and set development.

Greg Smith

Hi Greg, hi guys,

my first post in here.

As you guys are talking about the auto-stuff, maybe this is interesting for the "paint weights indicating quad density" thing http://people.csail.mit.edu/ibaran/autorig.pdf. Actually this is auto-rigging-skinning, considering this being strongly related to mesh density, it might be of worth taking a look into the algorithm (solver for heat equilibrium over the surface).

Yeah, automatic stuff is cool as long it's not ruining my work;-) Pretty sure automatic is going to change a lot but not right tomorrow. As far as I can see auto-retopo still needs my helping hand and my time.

It's cool that you guys received these positive responses at the conference, 3d-coat deserves it! Very nice piece of software indeed.

About this guy you talked to... It's just one guy! I think china is a weired idea. Im my eyes making a game is strongly depending on communication and how to make your team share the same vision. Don't think it's going to be easier or cheaper with china at all. Making a fantastic game needs a fantastic team! Most likely he will burn more money explaining his vision/communicating/reworking stuff than saving with cheaper artists. He is not the first to try this and not the first to fail. Maybe a thing like this makes sense when producing solely for one region like china or asia in general or just with a dedicated chinese team with it's own project. IMO outsourcing makes sense when you need specialists or when knowing you're going to f**k up the deadline without a hired helping hand.

[kollatt]

perhaps 3D-coat can brand this feature of andrew's "autopology" to show that it's the first to do so and for easier reference to the feature. i could maybe make a custom logo for the feature.

[philnolan3d]

perhaps 3D-coat can brand this feature of andrew's "autopology" to show that it's the first to do so and for easier reference to the feature. i could maybe make a custom logo for the feature.

I don't know about that, it sounds like apology, like they did something wrong and need to say they're sorry.

[Javis]

Yes it will still convert to voxels first. Hopefully this will change.

Ahh, you're right! No joy...

[kollatt]

pft. yeah.. apologize for being so awesome.

[kollatt]

man and i thought i was being original. i just thought autopology worked since the words were easily smooshed together. sounded clever x.x; man with the ability to retopo a basic mesh instead of voxels and options for placing poles this feature will be so fantastic. studios would be forced to buy 3D-coat if it can get the refining it needs. i don't think i've seen a feature in this program cause waves like this since voxels were introduced.

is there a high-res image of the 3D-coat logo for doing mockups with floating around?

[Thomas P.]

When will 3.3.09 be available for OSX users? The link still gives me a 3.3.08 download.

Thomas P.

[Andrew Shpagin]

When will 3.3.09 be available for OSX users? The link still gives me a 3.3.08 download.

Thomas P.

It is not easy to compile it for OSX (because of new libraries used), but I want to do this for 3.3.10 (Sunday or Monday).

So, it will be soon.

[Thomas P.]

It is not easy to compile it for OSX (because of new libraries used), but I want to do this for 3.3.10 (Sunday or Monday).

So, it will be soon.

Thank you Andrew. Happy to hear that as I would like to try out the new topology tools.

Thomas P.

[philnolan3d]

I tried the new update. I don't think I like the wizard function to be honest. When I hear automatic I think "fast" and "one click". Now obviously there can be more than one click, but in 3.3.09 it was actually possible to have a finished retopo in one click but with the wizard it is not. And because of so many popup windows it feels slower than it really is.

[fibric]

is it possible for osx user to add growl support for long time running tasks?

i played around with some of my old stuff created with blender and imorted it, done some testing tools exported it, but it took much time so minimized 3dc but missed the point where 3dc was done. it would be really really cool if 3dc would use growl notifications.

and ofc growl is also available for windows.

[kay_Eva]

I think the 3dCoat interface just needs some more love. But not sure if Andrew would ever have the time.

There are a ton of options. And options generally don't have a real time preview of what they mean which adds to the confusion level especially when there are like 10 toggles waiting to be clicked on. I think some sort of real time feedback preview of what these options will do to your objects would be helpful. I know I don't use 80% of buttons in 3dcoat mainly because I dont' know what they do.

And since most big operations take 10-20 seconds to complete, I know that if I just test out what they do, that it will take quite a while to see sometimes because of the processing times.

[Andrew Shpagin]

Updated to 3.3.10 (beta) (Win only now, I am trying to compile Mac/Linux but have some problem, anyway it will be solved soon)

- I brought auto-retopology to easily usable stage.It loors like wizard with several stages to draw guides, paint dense areas.

- Possibility to autoretopo external objects right from File->Import menu. It will automatically close holes, convert to voxels and perform autoretopology. On this stage I will make short pause (several days) to fix bugs and solve urgent issues from the forum.

- I have add new very important option to export setting "Perform coarsing". It is inverse to subdivision. If you will export low poly mesh with this option 3D-Coat will "sharpen" mesh in so way that in your 3D-package you will get same model as mid-poly in 3D-Coat after subdivision with Catmull-Clark. It is absolutely important if you want to get displacement and low poly mesh as result. If you will export displacement and low poly mesh, import in your package, subdivide mesh, apply displacement and get same result as in 3DC.

- New option in baking dialog - "Don't snap subdivided vertices" to get very smooth mesh on layer 0 and good displacement on other layers.

- New option in import dialog - "Don't center bounding box" to avoid loosing symmetry of imported object if it is only partially symmetrical.

- fixed: Curves in "E" panel will not be lost if you will try to edit them and drag parameters message box

- While resolving the issue http://bit.ly/c8FyEO I made 3DC to work correctly with Paint.net and other psd-painters.

- Fixed urgent problem mentioned by Tinker - http://bit.ly/9YOSsf (problem with incorrect merging to transformed volumes in some cases)

- fixed: Clear seams and Auto seams will act only on current UV set in UV room. Seams will be auto-set between faces in different UV sets.

- fixed laggy work of lines mode int "E" panel. Undo/redo will work correctly too.

- fixed bug in retopo room when it was impossible to drop invisible layer to the trash can.

[michalis]

Don't snap subdivided vertices

Thats interesting, waiting for mac - linux builds. This symmetry issue there, what about this? Not a bug but an issue alright.