3D-Coat 3.3 updates thread

[mocaw]

I am not some freakish fan, but I do have eyes. While yes, it is not like those perfectly calibrated examples shown from the sigg paper, it is very impressive still. Given that these features are almost alpha still, it is a bit early to pass judgement IMHO.

Did the sigg paper authors supply a working tool for us to try and break? Did they tailor their code to each model? There is so much work between a working example in a research paper and actually making a tool a regular end user can use that to compare them, so early, is nearly dishonest, and hardly fair.

I think the only thing fit to judge at this point is the coder, who has an amazing ability to take an abstract idea, and turn it into a useable tool. Thanks again Andrew, the votes so far are 99% with you and your developement of this feature!

[philnolan3d]

An experiment, take from it what you will. I took this hard edged shape and drew lines on every edge, trying to keep them fairly straight.

This is the result.

Here it is with no strokes at all.

[juanmanuel]

This looks like a very nice development. Now if Andrew could somehow detect hard edges in the auto-retopology routine... not that nice results aren't possible, as Phil has pointed out, but... tracing over every hard edge is unrealistic for some types of hard body models.

Anyway, things can only get better.

[rimasson]

This looks like a very nice development. Now if Andrew could somehow detect hard edges in the auto-retopology routine... not that nice results aren't possible, as Phil has pointed out, but... tracing over every hard edge is unrealistic for some types of hard body models.

Anyway, things can only get better.

+10

[Taros]

What we need is a possibility to retopo regular models too and not voxels only. At the moment you have to import objects to the voxel room and convert them to voxels. I hope Andrew will find a way to make it possible with non voxel objects too.

After some test, I must say, that this routine is a really big step forward. Ok, there are some points that could be optimised. I understand this. But for me personally, it is already a timer saver.

[polyxo]

A big help for quickly making guides on squarish objects was a hotkey-toggle inside the strokes tool

that makes it draw straight lines by picking a Start and an End-Point like in CAD-programs. Being able to

define circles by picking three points was cool too.

[Maximus3D]

Yes it's a must for this to work fluently with imported meshes or else it's pretty useless for a lot of people, me included. Because not all of us use voxelsculpting in 3DC and only some specific tools for certain tasks.

/ Magnus

[Tinker]

If you want to use this tool for the characters, the characters models that you would whant actually retop (I mean for project, at work) will different by detail from those which now show satisfactory results.

I'm not going to prove anything, and do not want to offend anyone. Just do not approve of unfounded optimism. Your conclusions should be based on facts and tests, rather than two connected primitives. And result with primitives while impressive only by potential.

(Do not think that I'm too serious)

[splodge]

What do you see in the attachment, Tinker?

[Taurus]

I don't think it's that useful, generated mesh is too high-poly, no room to tweak it, in the example with Orc it would take more time to fix 'auto-retopo' mesh than to make a new retopo in reto room because it's too high poly and edge flow is far from optimal.Same with cubes example - too high poly, you could do it faster and better with booleans.

Autoretopo will be useful if you can generate lower mesh resolution so it can be fixed fairly easy but you can do it right now in retopo room.You get too overexcited, it's nice new technology is getting into 3DC but I would rather see some basic stuff redone. Faster paint engine ? Raytraced AO bake ? More retopo tools ? Workflow improvements ? Instead dev time is spent on doing experimental tech which needs a lot of time and effort to tweak.

[Tinker]

What do you see in the attachment, Tinker?

Mediocre render with improperly configured intersection of water with glass.

[Javis]

Chill out a little.

This build you are using is very raw - I have seen some stuff directly from Andrew, that is much improved from the build you guys are using for the new auto retopo, and it is amazing! Greg, Don and myself have all seen it, and it truly is a amazing beauty. So sit tight guys, you wont have that build until after siggraph. Not to mention other features of auto retopo.

[Nemoid_]

interesting ! i agree about hard edge detecting and drawing.

also a workflow i would adopt is o choose a lower polycount, even when retopoing characters.

probably algorithm will do its job better, and if you need to edit the flow somewhere you wouldn't get mad. the rest can be done quickly adding manually more edge rows. i think artists will waste less time this way. i also think tke less point you start with the straighter some lines forcely will get. However , i'll wait the enhanced implementation Andrew will put after siggraph. I am sure it will end to be cool.

@Tinker : you're way too harsh. not a good attitude into a forum. if you don't like what you see remember its a w.i.p. and this is a beta.

p.s. advice for Andrew, after autoretopo, i would stop adding brand new features for something like 1 month, and instead try to empower 3D Coat making it smoother in performance, brush experience, workflows just to streamline workflows even more. this way no one could say its somewhat behind ZBrush. this work IMHO will benefit 3D Coat greatly 3D Coat has very nice things yet ebven in its UI which is more understandable than ZBrush one, so things can't go wrong at all.

p.s. 2 i've seen Newtek is going towards preproduction market. They have connection with virtual camera producers aa well (Intersense) this market will be a huge one. since 3D Coat connects so well with Lightwave, could be part of this process and it will be widely adopted in many studios. for Avatar pre pro they used many softwares, but primarily they used LW and Motionbuilder. So, i'd look for a tighter connection with those 2 apps first for sure.

[splodge]

Mediocre render with improperly configured intersection of water with glass.

Well that's not so bad, there's still hope for you. I was half expecting you to say that you saw the devil eating babies.

[Tinker]

By the way, I'll tell you how I became harsh. Every time I tidy up my hand from a mouse / tablet to switch the camera view with numpad I become more mad.

For the year waiting requested "viewport auto fit side views with rotating holding shift as in Z-brush" one hour easy to implement feature.

3D-Coat already detects approximate front/right/bottom view, showing it on the right corner of the gui. So, is it so hard to:

if ((shift && alt && LMouse)down && GUI "right")
   {
   CameraSwichRightView();
   }

or something like.

[geo_n]

What we need is a possibility to retopo regular models too and not voxels only. At the moment you have to import objects to the voxel room and convert them to voxels. I hope Andrew will find a way to make it possible with non voxel objects too.

After some test, I must say, that this routine is a really big step forward. Ok, there are some points that could be optimised. I understand this. But for me personally, it is already a timer saver.

+1

This would make 3dc ahead of topogun, etc. I'm sure Andrew just needs to code it to work in the retopo room.

Btw anyone tried to reduce the poly from autoretopo by doing this trick in Zbrush

http://www.newtek.com/forums/showpost.php?p=984037&postcount=84

[polyxo]

Btw anyone tried to reduce the poly from autoretopo by doing this trick in Zbrush

http://www.newtek.com/forums/showpost.php?p=984037&postcount=84

Yes. Didn't work with my examples.

Reconstructing subdivisions only works when a lower SubD-Level can still be created without triangles.

[LJB]

By the way, I'll tell you how I became harsh. Every time I tidy up my hand from a mouse / tablet to switch the camera view with numpad I become more mad.

For the year waiting requested "viewport auto fit side views with rotating holding shift as in Z-brush" one hour easy to implement feature.

Right that is true, This feature (which has been much requested for long time now) would surely fit with new tools and current expectations of the software.

I would really like to have two pence in here now. From a character point of view it can be used to save time, but like all things you really need to know where the flow needs to go and how things need to be animated. but even then dont expect anywhere near a complete mesh from it. For my Slinx video I would only choose to use the body section, not feet or head as the local detail is no way close to what is desired and even then I find i have to remove half and re sym. But for roughing out a general body as long as the guides are placed correctley and not hotch potched in a good flow is acheivable but correction is required.

Currently everything is far too avaeraged in size with too much detail being lost because of it. so local density must be looked at (i think this could prove very challenging however, But it depends on what Andrews Goal for this technology is).

If I might make a suggestion as to how i feel this tech would better fit, Currently the guides are guides and that is all. Now would it not be desirable to have this routine work to Fill the gaps SO say we work all of the complex areas and use this tech to go in and make a bit more sense of joining things together, rather than trying to reinvent the wheel with a completely automated solution that yields a 40% desirable complete mesh that needs complex correction.

Think about it carefully, and by example let me explain my logic (and this is purley based on character work, so Tinker this is probably irrelavent to your needs) so currently the way I work is to retopologise the more complex areas first Head, Hands, Feet then work to the bulk areas body limbs. That way I can be sure the number of loops is not effected and the flow is desirable on these more complex areas (For instance say I hvae a loop comprising of 12 polys around the wrist with the hand coming from those 12 polys its form is perfect for Low res cage). With the current complete solution I have no way of controlling the amount of poly's that run around the wrist area which means if I want a specific number of polygons in thos areas I would then have to chase the loops back to the trunk of the character in order to reduce or increase. and putting loops in or removing them takes just as much time as working them manually in the first place.

So if all those complex areas can be worked prior and the tech be used to match up and fill in bulk areas using the guides to control the flow would that not be the best Ideal solution?

Too much currently relys on luck and Polygon size to make this workable and the fact it produces Ngons and Spirals down tubular volumes meaning more flow chasing is very dissapointing. Please take this stuff as constructive. If we do not get these points across much time is wasted.

I am impressed with what it acheives but Its still a magpie feature in its current state. More refinement is essential to make it work and yield workable results. But I guess we all know this.

Anyway that is my two pence.

[Taurus]

Zbrush reconstruct works only with quad only mesh as far i know

[LJB]

Yes. Didn't work with my examples.

Reconstructing subdivisions only works when a lower SubD-Level can still be created without triangles.

Ah but its deeper than that Im afraid. It also relys on the amount and placement of these quads. You should never author a mesh in pure quads and expect to be able to reconstruct in ZBrush, you will find it breaks Very often.

I could ellaborate but I really cant be bothered. can we just take it that it does not work unless you are very careful as to how you create the mesh. and it should alway be initially created at its lowest resolution anyway. In the example posted above it is highly likely that that model was first created in quads at a lower resolution in the first place then the lower resolutions removed. therefore the lower quad structure was still attainable. If you need further explanation please experiment with ZBrush and try things for yourself.

[splodge]

yes, when reconstructing lower subdivisions in ZBrush you risk losing key features like your characters' nipples. ZBrush has no nipple detection technology in place and instead it just removes every nth polygon indiscriminately.

[Tinker]

Thanks, LJB, i understand your position. Retopo no more concerns me and I will not discuss it.

But you must understand and my position too. Quietly and politely stated position. What is useful and major i got from last half-year of updates? Critical bugfixes, discrete moving and multicore merge. Very nice but obvious things, I think. Compare to the previous "golden age" of true usefull updates its nothing.

I have great respect for Andrew's and the Team. Andrew fixed some bugs in normal baking critical for me in very short time. But are those bugs critical only for me? I think not. And you, guys, do not even discovered those bugs after a few months of use.

Are so many of you really use ptex? I mean not to play with inside 3D-Coat, but use. Who of you got renderman, the only tool support it, installed? Andrew spent half a year on it. And half a year work on retopology waiting ahead. While the request list full of features that would accelerate, simplify and make your work easier. Things you can do in minutes, not months. That's what I meant.

I have nothing more to say. Thank you

[LJB]

yes, when reconstructing lower subdivisions in ZBrush you risk losing key features like your characters' nipples. ZBrush has no nipple detection technology in place and instead it just removes every nth polygon indiscriminately.

No it does not remove anything. The structure of the mesh needs to be that that it will support a lower level. The step is not destructive at all (In zTool terms). All it does it look to see if the mesh will support a lower level and if the structure allows it simply drops the mesh down taking 4 existing quads and unifying. that is all. The detail would only be lost if you then deleted the Higher levels. Trouble is it would be extremely tricky to author a mesh with this lower level in mind. far simpler to just create at lowest. I get what the point of bringing it up was in the first place but currently things are not to a level that that level of finer control allowed (Not sure they will ever get there)

Anyway I too have said my tuppence. Tinker respect my freind. I am in 100% agreement over the little things. Nuff said.

[splodge]

No it does not remove anything.

I'm aware of how it works.

I was talking within the context of what some folks on this forum plan to use it for. It seems they're hoping to take their high poly models into ZBrush and reconstruct a lower subdivision and delete the higher subdivision. So using it this way it will result in key features being lost.

[Taurus]

Thanks, LJB, i understand your position. Retopo no more concerns me and I will not discuss it.

But you must understand and my position too. Quietly and politely stated position. What is useful and major i got from last half-year of updates? Critical bugfixes, discrete moving and multicore merge. Very nice but obvious things, I think. Compare to the previous "golden age" of true usefull updates its nothing.

I have great respect for Andrew's and the Team. Andrew fixed some bugs in normal baking critical for me in very short time. But are those bugs critical only for me? I think not. And you, guys, do not even discovered those bugs after a few months of use.

Are so many of you really use ptex? I mean not to play with inside 3D-Coat, but use. Who of you got renderman, the only tool support it, installed? Andrew spent half a year on it. And half a year work on retopology waiting ahead. While the request list full of features that would accelerate, simplify and make your work easier. Things you can do in minutes, not months. That's what I meant.

I have nothing more to say. Thank you

He has a point, many user requested features are being ignored.There is some imbalance there.Some people ask for Ptex they get it,some people ask for Auto-Retopo they get it.In the meantime whole rest is struggling with slow paint engine,poor AO baking,workflow issues ...

There is too much hype over auto-retopo, what I saw has very limited usability.Orc example - skin tight, grid-mesh it will be easier and faster retopo it from scratch in retopo room where you have control over polygons.Low poly objects - same, faster and easier to model from scratch or retopo.Looks like you are over optimistic.There is no way auto-reto will do any decent edgeloops for characters so you are left with organic and low poly objects, in the end it may happen you can do it same fast in reto because you won't waste your time on fixing.There is also more control over polys and loops.

Show me this auto-generated retopo and then fix to useble state, let's see if it's indeed faster than manual retopo taking into consideration fixing time.

[haikalle]

The fact is that Andrew can't do all these features requests at once. He has to do it one by one. Sometimes the order is not what we would like

but sometimes it goes like that. More and more Ptex support is coming around. Just wait and see. Judging new-generated retopo system is just too soon.way too soon. Wait for week or two. Andrew has already said that it's still raw and gets better and better. There is already tons of new features in 3d-coat. There is a lot of playground to mess with. We don't always have to wait the latest update to do some amazing art with 3d-coat.

He has also said that the paint engine gets some improments soon. It will come...

[polyxo]

Thanks, LJB, i understand your position. Retopo no more concerns me and I will not discuss it.

But you must understand and my position too. Quietly and politely stated position. What is useful and major i got from last half-year of updates? Critical bugfixes, discrete moving and multicore merge. Very nice but obvious things, I think. Compare to the previous "golden age" of true usefull updates its nothing.

I have great respect for Andrew's and the Team. Andrew fixed some bugs in normal baking critical for me in very short time. But are those bugs critical only for me? I think not. And you, guys, do not even discovered those bugs after a few months of use.

Are so many of you really use ptex? I mean not to play with inside 3D-Coat, but use. Who of you got renderman, the only tool support it, installed? Andrew spent half a year on it. And half a year work on retopology waiting ahead. While the request list full of features that would accelerate, simplify and make your work easier. Things you can do in minutes, not months. That's what I meant.

I have nothing more to say. Thank you

For my needs the recent changes, Ptex and Auto-Quadrangulation are very welcome.

Your posting only represents a very Game-Creation centric point of view.

But there's many users who don't work in this Industry and couldn't care less

for Games and Character-Modeling who also would like to use this program.

Concerning Ptex: I have no problems to render with maps from 3DC with other engines

and am glad that I don't have to think about UV's at all.

[Nemoid_]

Maybe Andrew focuses too much on new technologies to implement, to make 3D Coat more attractive, and i can understand these reasons.

However, as i said if he puts some effort on solving 3D Coat bugs and provide an even faster workflows this will give to the app even more value than most innovations.

3D Coat is innovative yet. It has voxels technology which is great, good manual retopo tools and upcoming autoretopo to help doing this job even better, painting layers with fusion methods, ptex and more. if all this will be optimized to give users an even more pleasant experience with what's already there, the app will be welcomed very well among pros and studios. That's what makes the difference in sellings too. Look at Mari. comes from developing fro WETA , everyone wants it even if they dunno if it will be so gret (I hope so btw)

[splodge]

3D Coat's success hangs on quadrangulation.

The only thing that matters in 3D is the speed at which you can get something out of your brain and onto your hard drive.

If people can box model quicker than they can sculpt a model in voxels AND retopo that model then they're always going to ignore 3D Coat.

So you all shutup already and show this new feature some love or else I'm going to send one of these to all your PM inboxes..

[haikalle]

I think it was good plan to make PTex, multi-threaded and auto retopo system. These are good selling point in the long run. When the next major updates comes out where

we have to pay for it, after that Andrew has more time to make it more faster under the hood.